Stock Analysis · Bulls vs Bears

Broadcom Inc. AVGO

🐂 Bulls vs Bears 🐻Report #312 Oct 2026
★ Willem owns this stock · 2.8% of the portfolio
34
moderate risk▼ −18 vs last
🐂 9vs10 🐻
Bears lead
40%
▲ Chance of the biggest tailwind the forecasts miss
30%
▼ Chance of the biggest headwind the forecasts miss
The verdict

The operating evidence still leans strongly bull: Revenue has beaten guidance in 8 of 8 quarters. Free cash flow over the last twelve months was $39.4B. Net debt is only ~0.6x EBITDA. RPO (contracted future revenue) is $179B, and supply is pre-bought for FY27–28. What has moved the wrong way is the quality of that growth: It increasingly runs through two cash-hungry AI labs that Broadcom itself helps finance. The…

The Bulls vs Bears podcast is being made
Two AI hosts debate this report, NotebookLM-style. It appears here automatically within a few hours.

Written by Claude (Anthropic’s AI) from public sources · Willem owns AVGO · AI can make mistakes — numbers in brackets point to the sources · Not financial advice

🐂 Bull arguments 9

  1. 1Eight straight quarters of beating its own guidanceStrong
  2. 2$39B of free cash flow a year on a light balance sheetStrong
  3. 3Still the default partner for custom AI chipsStrong
  4. 4It sells the network for every AI cluster, not just its ownStrong
  5. 5Pre-bought supply through 2028 is a moat rivals lackMedium
  6. 6The stock trades below its own five-year multipleMedium
  7. 7Financing the labs locks in the two biggest new customersMedium
  8. 8Apple to 2031 and a recovering non-AI chip baseMedium
  9. 9VMware still throws off near-pure-margin cashMedium

🐻 Bear arguments 10

  1. 1Broadcom guarantees its customers' debts, and won't size the totalHigh
  2. 2Growth is concentrating in two loss-making AI labsHigh
  3. 3Management's story on racks, margins and guarantees keeps shiftingMedium
  4. 4The FY27–28 guides imply far fewer gigawatts than the pitchMedium
  5. 5Gross margin keeps falling as memory costs pass throughMedium
  6. 6Google is building real second sources for its TPUMedium
  7. 7Software growth is mostly an accounting changeMedium
  8. 8$126.8B of purchase commitments meet physical bottlenecksMedium
  9. 9Taiwan, China and a federal label on its top customerLow
  10. 10A 74-year-old CEO, no successor, and insiders sellingLow

Bull arguments show how well they survive the devil’s advocate (strong / medium / weak); bear arguments show the damage they could do after the counter-argument (high / medium / low).

Bull case

🐂 Bull #1Eight straight quarters of beating its own guidance

  • The point: Broadcom's guides are conservative and reliable. That makes the FY27 ~$115B AI number more credible than a typical multi-year promise.
  • Evidence:
    • Revenue beat the company's own guide in all eight quarters from Q4 FY24 to Q3 FY26, by 0.4% to 3.5% [2,4,5,6,7].
    • AI revenue met or beat its guide every time:
      • Q4 FY24: >$3.5B guided, $3.7B delivered.
      • Q3 FY26: $16.0B guided, $16.7B delivered (+4.4%) [7,8].
    • The margin guide (EBITDA, then operating margin) was beaten by about 1 point in all eight quarters. Q3 FY26: 67% guided vs 67.9% actual [2,6].
    • The FY27 AI outlook rose from "in excess of $100B" (March/June) to "approximately $115 billion", with "secured the supply" (2026-09-02) [8,10,11].
    • PwC audits the company. No restatement or material weakness was found. The 10-Q says controls are effective [1,3].
  • Why it matters: a manager who under-promises lets the market price the guide close to face value. That makes the FY27 plan the base case rather than the bull case.
  • 😈 Devil's advocate:
    • Revenue reliability is not the same as margin reliability. In March Tan told an analyst modelling rack dilution that he "must be a bit hallucinating", and gross margin then fell 4 points in two quarters [11,8].
    • The FY28 ~$230B is described as "line of sight", not "secured". Tan says he does "not intend to update it on a quarterly basis" [8].
    • Past beats came when the customers paying were cash-rich hyperscalers. FY27–28 depends on labs that need outside financing.
  • Verdict: Strong On revenue, the record is as clean as any in large-cap semis, though it says less about the newer, lab-funded part of the plan.

🐂 Bull #2$39B of free cash flow a year on a light balance sheet

  • The point: Broadcom funds the AI ramp, the dividend and debt paydown from its own cash flow, with no need for equity.
  • Evidence:

    • Free cash flow over the last four quarters [2,4,5,6]:

      Quarter FCF Share of revenue
      Q4 FY25 $7.47B 41%
      Q1 FY26 $8.01B 42%
      Q2 FY26 $10.26B 46%
      Q3 FY26 $13.67B 46%
      Total (LTM) $39.4B

    • Balance sheet at 2026-08-02 [1]:

      • Cash $23.98B, plus an undrawn $7.5B revolver and $0 commercial paper.
      • Debt principal $61.1B, so net debt is $37.1B, about 0.6x LTM adjusted EBITDA ($61B, estimate).

    • Debt terms:

      • Fixed coupon averages ~4% with ~7.4 years to maturity [8].
      • $0 falls due in FY27. Only ~$7.4B falls due through FY28 [1].
      • In FY26 Broadcom has repaid or repurchased $10.5B of debt [1].
  • Why it matters:
    • If a lab slips, Broadcom will not be forced into dilutive funding.
    • It can absorb a mid-size guarantee loss from one year's cash flow.
  • 😈 Devil's advocate:
    • Part of the cash flow is flattered by working capital. Receivables nearly doubled ($7.1B → $13.7B) while $3.95B of receivables were factored (sold to banks) in nine months [1].
    • The ~$29B 10-Q backstop is bigger than the cash pile [1].
    • $52.7B of purchase commitments fall in FY27 alone [1].
  • Verdict: Strong The cash generation is real and the debt profile is long and cheap. The caveat is that the guarantees sit outside these numbers.

🐂 Bull #3Still the default partner for custom AI chips

  • The point: every big buyer that wants a cheaper alternative to Nvidia GPUs ends up at Broadcom for at least one programme.
  • Evidence:
    • Six XPU customers. Tan says "four of them are just simply going to be huge" [9].
    • Contracts:
      • A Google Long-Term Agreement for "future generations" of TPUs, plus supply assurance through 2031 [13].
      • Apple multi-generation custom silicon through 2031 [14].
    • Volumes guided:
      • Anthropic: 5 GW of TPU 8i in 2027 and ~10 GW more in 2028.
      • OpenAI: 1.3 GW in 2027, line of sight to more than 5 GW in 2028.
      • Meta: ~3 GW through 2028 [8].
    • RPO is $179.2B, ~25% of it recognised within 12 months. It includes a long-term custom-accelerator contract signed in Q2 [1].
    • Counterpoint projects ~60% design-partner share in 2027 [47].
    • Tan says competitors' all-in cost is ~$40B per GW and Broadcom's is "less than half" [9].
  • Why it matters: design wins run for several chip generations, and the custom-chip market is growing faster than GPUs.
  • 😈 Devil's advocate:
    • The customer owns the design and is adding second sources (Bear #6).
    • Google's agreement promises a relationship, not exclusivity. No generations are named [13].
    • The "six customers" headline hides how dependent the plan is on two labs (Bear #2).
  • Verdict: Strong / Medium. The franchise is intact, but share will drift down from today's peak.

🐂 Bull #4It sells the network for every AI cluster, not just its own

  • The point: Broadcom's switch chips (Tomahawk, Jericho) and network cards go into GPU clusters as well as XPU clusters. That spreads the business beyond the six custom-chip customers.
  • Evidence:
    • AI networking revenue was $4.52B in Q3, up more than 2.5x year on year. Tan expects it to grow as fast as XPUs [9,50].
    • Tomahawk 6 (102.4T) is deployed in "pretty much all" hyperscalers building XPUs with Broadcom [9].
    • Tomahawk 7 (204.8T) "just taped out" (Q3 call) [9].
    • At the OCP Summit (release 2026-10-08) Broadcom showed Tomahawk 6, Tomahawk Ultra, Jericho 4, Thor Ultra network cards and third-generation co-packaged optics, with 15+ partners [48].
  • Why it matters: networking is less concentrated by customer and harder to multi-source than an XPU.
  • 😈 Devil's advocate:
    • IDC ranked Nvidia the #1 data-centre Ethernet switch vendor in Q1 2026: 21.5% share, revenue +193% y/y, just ahead of Arista at 20.7% [49].
      • Nvidia's Spectrum-X does not use Broadcom silicon. Arista and the white-box makers largely do.
    • Scale-up is still contested between NVLink, UALink and ESUN/SUE-T. Broadcom has reportedly walked away from UALink [50].
  • Verdict: Strong / Medium. Ethernet is winning, but Nvidia now takes a big slice of the Ethernet win with its own silicon.

🐂 Bull #5Pre-bought supply through 2028 is a moat rivals lack

  • The point: when packaging, HBM and lasers are the bottleneck, customers pick the partner who can promise delivery.
  • Evidence:
    • Purchase commitments are $126.8B: $52.7B for FY27 and $73.0B for FY28 [1].
    • On FY28, Tan said "we have secured the supply" [8].
    • Broadcom is building a substrate fab in Singapore (deploying from FY27) and more than tripling laser capacity [8].
    • MediaTek's board had to approve a ~$5B budget to lock in its own supply (previous report).
  • Why it matters: this turns contracted demand into shippable revenue, and second sources struggle to get capacity.
  • 😈 Devil's advocate:
    • The commitments are fixed while demand rests on lab financing. If a ramp slips, Broadcom holds the inventory (Bear #8).
    • Nvidia's supply commitments are far larger.
  • Verdict: Medium It is a moat in a shortage and a liability in a slowdown.

🐂 Bull #6The stock trades below its own five-year multiple

  • The point: after a ~25% fall from the June record, the stock prices FY27 earnings more cheaply than its history, so expectations are no longer stretched. (New argument.)
  • Evidence:
    • Price $361.54 (2026-10-09) vs the $481.57 record close (2026-06-02) [33,34].
    • Forward P/E is ~20.9x, vs a FY21–FY25 average of ~24x and 44x in FY25 [34].
    • FY27 consensus EPS is $19.39, about 18.6x at today's price [33].
    • Short interest is 1.1% of float (2.0 days to cover) [34].
    • The average analyst target is $531 (40 Strong Buy, 7 Buy, 3 Hold, 0 Sell) [33].
    • After Q3, Cantor, Rosenblatt, BMO and Bernstein had targets of $575–600 [36,33].
  • Why it matters: if FY27 is delivered, the multiple no longer has to expand for the stock to work.
  • 😈 Devil's advocate:
    • Trailing EV/EBITDA is 33.7x [34].
    • Consensus already contains the full ~$115B AI path, and FY28 consensus reportedly sits near Tan's "$30+" EPS [35]. The base case is priced; only the upside isn't.
    • The multiple fell for a reason: the financing layer.
  • Verdict: Medium The stock is cheaper but not cheap, and the discount pays you for a real risk.

🐂 Bull #7Financing the labs locks in the two biggest new customers

  • The point: by arranging and partly guaranteeing financing, Broadcom becomes hard to replace at Anthropic and OpenAI. The convertible notes also give equity upside.
  • Evidence:
    • Anthropic's five-year TPU lease commitment is $125.2B. Its Broadcom equipment leases total ~$161.2B and are largely non-cancellable [20,22].
    • Anthropic has placed cash in a restricted account for Broadcom's benefit [20].
    • The up-to-$42B notes can convert into Anthropic equity. Anthropic is reportedly targeting a ~$2T IPO valuation [21].
    • The XPV platform is designed for more than 20 GW through 2028 [1].
  • Why it matters: it turns weak lab credit into financing labs can afford, which brings gigawatts forward. If the labs succeed, Broadcom owns part of the upside.
  • 😈 Devil's advocate:
    • This is the same structure critics call "circular financing", and Anthropic's filing flags "potential conflicts of interest" [20].
    • It only works while credit markets stay open (Bear #1).
  • Verdict: Medium It is the bull and bear case at once.

🐂 Bull #8Apple to 2031 and a recovering non-AI chip base

  • The point: the legacy chip business has stabilised and its biggest customer is signed up for years.
  • Evidence:
    • Non-AI semis were $4.2B in Q3 (+5% y/y), with $4.3B guided for Q4 [8].
    • In the last downturn non-AI semis fell 23% y/y (Q4 FY24) [8].
    • Apple's multi-generation custom silicon agreements run to 2031 [14].
  • Why it matters: this removes a tail risk on ~15% of revenue.
  • 😈 Devil's advocate:
    • The Apple 8-K gives no dollar value [14].
    • Wireless is still declining.
    • Non-AI is now too small to cushion an AI slowdown.
  • Verdict: Medium / Weak.

🐂 Bull #9VMware still throws off near-pure-margin cash

  • The point: customers can't quickly rip out VMware, and the software arm earns 94% gross and ~84% operating margin.
  • Evidence:
    • Software revenue was $8.75B in Q3 (+29%) [2].
    • Software gross margin 94%, segment operating margin ~84% [8].
    • Most new software contracts are now non-cancellable [1].
  • Why it matters: about $35B a year of software revenue funds everything else at software-company margins.
  • 😈 Devil's advocate: stripping out upfront licences, software grew only ~8.6% (Bear #7), and recurring revenue growth (ARR) slowed from 19% to 15% in three quarters [8,10,11].
  • Verdict: Medium (weakened again). The cash is real, but the growth narrative is mostly timing.

Dropped since last report: "The CEO is tied in through FY30". The re-test confirmed his award maxes out at $120B of AI revenue, which management's own guide already exceeds. It works as retention only and is not a bull argument; it now sits under Bear #10.

Bear case

🐻 Bear #1Broadcom guarantees its customers' debts, and won't size the total

  • The point: Broadcom has become a credit provider to its biggest customers, and the exposure is growing faster than the disclosure.
  • Evidence:
    • The 10-Q (Note 10, period to 2026-08-02) describes a backstop on the first XPV tranche [1]:
      • Broadcom pays "any difference between 85% of the outstanding amounts owed… and the value of the AI racks" if the customer defaults.
      • The maximum is ~$29B undiscounted. Fair value is "not material". Nothing has been paid.
      • Separately, the customer may issue Broadcom up to $42B of convertible notes. None have been issued.
    • The 10-Q also reserves room for more: "Where necessary, we may provide residual value guarantees" [1].
    • Since the 10-Q:
      • A further $60B has gone into syndication: a $42B senior Class A with a partial Broadcom residual-value guarantee, and an $18B junior tranche [23].
      • IFR describes the senior debt as two $21B loans maturing 2033-09-30 [24].
      • Talks for OpenAI are reported at ~$30B (Bloomberg, 10-09) or more than $50B (WSJ, 10-07) [25,26].
    • The total is undisclosed:
      • On the Q3 call the CFO said: "I can't give you an overarching look at like what's the max" [8].
      • BofA credit research models maximum residual-value-guarantee exposure at $370B across 20 GW by mid-2029. In a full-default scenario it puts losses at ~$42B. BofA cut Broadcom's bonds to Marketweight/Neutral (2026-08-10/11) [28,29].
    • Credit-market and agency reaction:
      • S&P reportedly called the first tranche "credit negative" (~2026-06-11) without changing the A- rating [30].
      • 5-year CDS hit a reported record ~122bp in August [31].
      • Ratings last confirmed in a filing (January 2026): A3 / A- / BBB+, all on positive outlook [15].
  • Why it matters:
    • A lab default would hit Broadcom three ways at once: lost orders, a guarantee call, and losses on the notes.
    • The guarantee would also be called at the worst moment, when used AI racks are cheapest.
  • 🛡️ Counter-argument:
    • Losses only arise after the racks are resold, and Broadcom can take over the lease or sell the racks back "at a fixed price" [1].
    • The junior tranche absorbs losses first.
    • Even BofA's full-default loss (~$42B) is roughly one year of free cash flow.
    • Four of six XPU customers fund themselves [8].
  • Severity: High The size is manageable against cash flow, but the exposure is correlated with the revenue risk and it keeps growing without a disclosed cap.

🐻 Bear #2Growth is concentrating in two loss-making AI labs

  • The point: the FY27–28 plan rests more and more on Anthropic and OpenAI, whose ability to pay depends on raising outside capital.
  • Evidence:
    • Broadcom concentration [1,8]:
      • Top five end customers were ~55% of Q3 revenue, up from ~40% a year earlier.
      • One distributor handled 50% (32% a year earlier).
      • Tan says Anthropic will be the "largest XPU customer in 2027".
    • Anthropic, from Reuters and Fortune reporting on its confidential prospectus (2026-09-29 and 10-01) [20,21,22]:
      • 2025 revenue ~$4.6B and an operating loss ~$8.1B.
      • At least $518B of compute commitments, ~80% take-or-pay.
      • Q2-2026 adjusted operating profit $559M.
    • OpenAI:
      • On 2026-10-08 it said annualised revenue was ~$50B, against ~$68B reported earlier, which included partner gross revenue. AVGO closed −4.4% that day [27].
      • Tan himself calls OpenAI's finances "less clear" than Anthropic's (previous report).
    • The April 2026 8-K makes Anthropic's 3.5 GW "dependent on Anthropic's continued commercial success" [13].
  • Why it matters: a two-quarter funding gap at one lab moves tens of billions of Broadcom revenue and also triggers Bear #1.
  • 🛡️ Counter-argument:
    • Both labs are growing very fast:
      • Anthropic's run-rate is reported above $47B in May.
      • Anthropic raised $65B in May; OpenAI raised $122B in March [22,62].
    • Anthropic turned an adjusted operating profit in Q2 [21].
    • Google, Meta and Apple are still large and fund themselves.
  • Severity: High Concentration rose in a single year, and the two customers it now rests on have the weakest balance sheets.

🐻 Bear #3Management's story on racks, margins and guarantees keeps shifting

  • The point: three times in a year, management's verbal framing was contradicted by later filings or results. That makes the unverifiable parts of the pitch (content per GW, guarantee size) harder to trust. (New argument.)
  • Evidence:
    • Racks:
      • December 2025: "It's a real system sale" [12].
      • 2026-06-03: "No racks. It is all a chip business only." Asked about Anthropic, Tan said the deal "was not backstopped in that sense" [10].
      • Days later the XPV platform launched with a backstop. The 10-Q describes the partner buying "AI racks", and a new risk factor mentions leasing "full AI racks or systems" [1,30].
    • Gross margin:
      • 2026-03-04: Tan said "We will not be affected by the gross margin by more and more AI products" [11].
      • Non-GAAP gross margin then went 77.1% → 75.0% → ~73% guided [2,8].
    • Content per GW drifted from "not far from" ~$20B (Q1) to "$20 billion to $30 billion" (Q3) [11,8].
    • Guarantees: the CFO called them "modest" and "low risk" but declined to give a maximum [8].
  • Why it matters: investors are increasingly asked to trust numbers Broadcom does not disclose.
  • 🛡️ Counter-argument:
    • The filings, which carry legal liability, did disclose the backstop, the notes and the margin effects promptly [1].
    • Revenue guidance, the thing management controls most, has been reliable (Bull #1).
    • The June call may have pre-dated the final XPV terms.
  • Severity: Medium It is a credibility discount, not evidence of misstatement.

🐻 Bear #4The FY27–28 guides imply far fewer gigawatts than the pitch

  • The point: the dollar guides and the per-gigawatt claims don't fit together. Either many announced gigawatts land after FY28, or Broadcom's share of each gigawatt is lower than claimed.
  • Evidence:
    • Q3 call: Bernstein's Rasgon divided the guides by the ~10 GW (2027) and ~20 GW (2028) plans and got $11–12B per GW. Tan did not dispute the arithmetic. He said not all ~30 GW reach production by FY28, and that he has "a pretty high degree of confidence" in ~$350B of FY27–28 shipments [8].
    • In the same call Tan put Broadcom content at "$20 billion to $30 billion per gigawatt… relatively stable" [8].
    • Interpretation: at $25B/GW, $350B equals ~14 GW over two years, versus ~30 GW of customer plans named on the call. That puts roughly half the named gigawatts beyond FY28, or makes content lower.
  • Why it matters: if gigawatts slip, FY28 is at risk. If content is lower, the per-GW pitch overstates Broadcom's take of each AI dollar.
  • 🛡️ Counter-argument:
    • Slipped gigawatts are FY29 upside, not lost revenue.
    • Tan calls the guide "judged". Its dollars have been beaten every quarter.
  • Severity: Medium (unchanged). It is a transparency issue that makes FY28 harder to underwrite.

🐻 Bear #5Gross margin keeps falling as memory costs pass through

  • The point: each XPU carries more high-bandwidth memory (HBM) bought from suppliers and passed through, so each dollar of revenue earns less gross profit.
  • Evidence:

    • Non-GAAP gross margin by quarter [2,4,5,6,8]:

      Q4 FY25 Q1 FY26 Q2 FY26 Q3 FY26 Q4 FY26 guide
      77.9% 77.0% 77.1% 75.0% ~73%

    • Semiconductor segment gross margin was ~67% in Q3. The CFO first said "76%" on the call, then corrected it [8].

    • TrendForce sees 2027 HBM contract prices up 70–140% and blended prices +121% [58].

    • The new 10-Q risk factor says AI racks and systems "will likely… compress or lower future gross margin" [1].

  • Why it matters: less cushion if XPU pricing pressure arrives from MediaTek. If HBM costs can't be passed through in full, operating margin follows.
  • 🛡️ Counter-argument:
    • Non-GAAP operating margin rose: 66.2% → 66.4% → 67.3% → 67.9%. Gross-profit dollars are surging [2,4,5,6].
    • Morningstar argues XPUs are operating-margin neutral (reported).
  • Severity: Medium It is harmless while operating margin holds at ~66%, and it becomes a problem below ~62%.

🐻 Bear #6Google is building real second sources for its TPU

  • The point: the anchor custom-chip customer is deliberately spreading work, so Broadcom's share of future TPU generations will fall.
  • Evidence:
    • MediaTek told its July call it has a two-generation AI-ASIC roadmap: generation one in mass production in Q4 2026, generation two in early 2028. It raised its 2027 share target to 15–20% [42].
    • TrendForce (2026-06-23, supply-chain report) says MediaTek won major TPU v9 orders after Google's 448G SerDes plan lagged. It adds that Broadcom "has reportedly lost its leading position for now" [43].
    • Marvell's Google warrant (8-K, 2026-08) covers up to 58.97M shares at $206.58, vesting per $500M of Google custom revenue, i.e. up to ~$120B through FY33. It covers TPU-attached chips, not the TPU itself [45].
    • Macquarie's June model saw Broadcom's TPU share falling from ~95% in 2026 to ~65% by 2028 [37].
  • Why it matters: each second source caps Broadcom's price and share on its largest single programme.
  • 🛡️ Counter-argument:
    • On the Q3 call Tan said Broadcom is "shipping the TPU version 8i ahead of the MediaTek version" [8,9].
    • Anthropic's 2028 ~10 GW is reportedly TPU 9 via Broadcom [50].
    • Macquarie re-upgraded on 2026-09-02, saying the risk is "largely reflected" in the price [37].
    • Which company makes which TPU 8 variant is still unresolved. Google names no partner, and a pre-Cloud-Next report says the reverse of the Broadcom call [40,41].
  • Severity: Medium Share erodes but dollars still grow, unless the v9 flagship moves.

🐻 Bear #7Software growth is mostly an accounting change

  • The point: the 29% software growth in Q3 came mostly from booking multi-year licences upfront, not from underlying demand. (New, quantified argument.)
  • Evidence:
    • Upfront licence revenue was $3,465M in Q3 vs $1,916M a year earlier [1]. That is 39.6% of software revenue, vs 28.2% a year earlier.
    • Excluding it, software revenue was ~$5,287M vs ~$4,870M, about +8.6% (our arithmetic from [1]).
    • The trigger: "we no longer have termination for convenience provisions for the majority of new software contracts" [1].
    • Q4 software is guided down quarter on quarter to ~$8.7B [8].
    • VMware ARR growth went 19% → 17% → 15% [8,10,11].
    • Legal and regulatory:
      • Tesco's UK claim (≥£100M) has a trial window of 2027-11 to 2028-02 [56].
      • CISPE's EU complaint is pending, with five trade groups asking for interim measures in July [55].
    • A critical vCenter flaw (CVE-2026-59310) was exploited in August, with 361 victim IPs observed [57].
  • Why it matters: if growth is single-digit, the 2027 renewal wave of the 2024 VMware contracts becomes a real test. Software is ~30% of revenue and most of the profit cushion.
  • 🛡️ Counter-argument:
    • Non-cancellable contracts lock in revenue and cash for years [1].
    • ~9% underlying growth on a 94%-gross-margin base is still good.
    • No churn has been disclosed.
  • Severity: Medium / Low.

🐻 Bear #8$126.8B of purchase commitments meet physical bottlenecks

  • The point: Broadcom has committed to buy the supply, but customers need power, land and buildings ready on time to take delivery.
  • Evidence:
    • Purchase commitments: $52.7B for FY27 and $73.0B for FY28 [1].
    • Inventory doubled to $4.5B in nine months [1].
    • Tan on customers' "land, power and shell": "more than a big concern". He also asked whether chips will be "deployed on a timely basis" [8].
    • A new 10-Q risk factor says customers may "underestimate or be unable to obtain… power or water capacity" [1].
  • Why it matters: a slip at one lab leaves Broadcom holding components and inventory.
  • 🛡️ Counter-argument:
    • Hyperscaler 2026 capex is ~$725B in aggregate and still rising. Amazon is at ~$220B [59].
    • Shortage conditions mean components can be redirected to other customers.
  • Severity: Medium

🐻 Bear #9Taiwan, China and a federal label on its top customer

  • The point: the geopolitical risks are spread across several fronts, and one now touches its future largest customer.
  • Evidence:
    • China: FT/Reuters (2026-09-23) report that SASAC, the state-asset regulator, surveyed Broadcom switch use in state data centres, where Broadcom may be ~90% of installed equipment. Informal guidance to cut back is possible [52].
    • Tariffs: Section 232 phase two (tariffs on imported chips and possibly servers) is still unsigned. Reports say the administration is less open to broad data-centre exemptions [53,54].
    • Anthropic: the D.C. Circuit upheld the Pentagon's "supply chain risk" label 2–1 (~2026-09-25). Anthropic is "considering all options" [51]. Broadcom's 10-Q added a risk factor saying it may have to suspend dealings with entities designated a supply chain risk [1].
    • Taiwan: ~95% of outsourced wafers come from TSMC (FY25 10-K) [3].
    • The US–China truce runs to 2027-01-10 [61].
  • Why it matters: each front is a margin or share nick except Taiwan and the label, which are low-probability, high-impact tails.
  • 🛡️ Counter-argument:
    • As reported, the label covers Defense Department work, not commercial chip sales.
    • US data centres are exempt from the phase-one tariff [53].
  • Severity: Low / Medium.

🐻 Bear #10A 74-year-old CEO, no successor, and insiders selling

  • The point: the strategy and the customer relationships run through one person, and insiders are net sellers into the AI ramp.
  • Evidence:
    • Hock Tan is 74 with no named successor. The board says succession is focused on "internal candidates" [16].
    • CFO Amie Thuener took over on 2026-06-12. On 2026-07-02 she adopted a 10b5-1 plan to sell 50% of her net vested shares [1].
    • Insider trades, 2026-04 to 2026-10 [18,19]:
      • Co-founder and director Samueli sold 654,241 shares at $382 on 2026-06-24 ($250M, pre-arranged plan) and 702,190 more on 2026-09-23 (price not found).
      • Chief Legal Officer Brazeal sold 75,000 shares in open-market sales at $379–401 ($29M, plan status not confirmed).
      • Directors Page and Delly sold small amounts.
      • The only open-market buy was director Harry You: 1,000 shares at $374 (2026-06-11).
    • Tan, Kawwas and Velaga: no Form 4 sales in the window [18].
  • Why it matters: a sudden CEO exit would hit the multiple. Sustained insider selling is a weak but real sentiment signal.
  • 🛡️ Counter-argument:
    • Tan's FY28–30 award keeps him in place to FY30 [16].
    • Samueli's sales are pre-planned and small against his holding.
    • Tan himself has not sold.
  • Severity: Low / Medium.

Tailwinds NOT in the forecasts

Common basis (estimate): FY27 consensus is revenue $173.9B and EPS $19.39 [33]. FY28 consensus is ~$270B and ~$30 EPS, from one paywalled source citing a Koyfin chart (low confidence) [35]. ~4.94B diluted shares and a ~16% non-GAAP tax rate (Q4 guide) [8]. AI incremental operating margin is assumed at ~55%, below the 61% semis segment margin because of HBM pass-through. That gives ~$0.094 EPS per $1B of AI revenue.

  • FY28 AI revenue reaches management's ~$230B, above what consensus appears to model — 40% chance by FY28 (Nov 2028)

    • Impact on revenue: +$15B (+5.6%) vs ~$270B consensus FY28

    • Impact on EPS: +$1.41 per share (+4.7%) vs ~$30 consensus FY28

    • How we got the number (estimate): consensus FY28 ~$270B minus ~$55B non-AI ≈ $215B AI → $15B gap × 55% × (1 − 16%) ÷ 4.94B ≈ +$1.41

    • Expected value: 0.40 × $1.41 ≈ +$0.56

    • What it is: the "line of sight" FY28 number delivered in full.

    • Why it's not priced in: the only consensus read we have implies ~$215B of AI revenue. Analysts haircut FY28 for lab financing.

    • Trigger to watch: the FY27 outlook at Q4 FY26 (2026-12-09). Any upgrade from "line of sight" to "secured" for FY28.

    • Why this probability: management has beaten every guide, but FY28 needs both labs funded and ~$230B squares with the per-GW gap only if many gigawatts slip.

  • A seventh large XPU customer is announced — 25% chance within 36 months

    • Impact on revenue: +$5B (+1.9%) vs ~$270B consensus FY28

    • Impact on EPS: +$0.47 per share (+1.6%) vs ~$30 consensus FY28

    • How we got the number (estimate): 0.5 GW in the first full year × ~$10B/GW guide-implied content × 55% × 0.84 ÷ 4.94B ≈ +$0.47

    • Expected value: 0.25 × $0.47 ≈ +$0.12

    • What it is: for example Microsoft Maia or an AWS programme moving to Broadcom.

    • Why it's not priced in: consensus models only named programmes. No report of AWS or Microsoft moving to Broadcom was found.

    • Trigger to watch: a new customer named on a call, or an 8-K long-term agreement.

    • Why this probability: Microsoft is reportedly with Alchip/Marvell and AWS packages Trainium itself, so a switch is plausible but not the base case.

  • Scale-up Ethernet and co-packaged optics become a second networking leg — 35% chance of material revenue by FY28

    • Impact on revenue: +$3B (+1.1%) vs ~$270B consensus FY28

    • Impact on EPS: +$0.28 per share (+0.9%) vs ~$30 consensus FY28

    • How we got the number (estimate): $3B incremental × 55% × 0.84 ÷ 4.94B ≈ +$0.28 (no market-size anchor; not quantifiable more precisely)

    • Expected value: 0.35 × $0.28 ≈ +$0.10

    • What it is: Tomahawk Ultra and SUE-T for scale-up inside racks, plus Davisson CPO.

    • Why it's not priced in: Tomahawk Ultra deployment "starting… this quarter and in FY27"; CPO has no disclosed revenue [9,48].

    • Trigger to watch: OCP (10-12 to 10-15) hyperscaler adoption. Tomahawk 7 products, guessed for late 2027 to early 2028.

    • Why this probability: the scale-up standards war is unresolved and Nvidia's NVLink is entrenched.

  • Capital-return step-up in December — 45% chance by 2026-12-09

    • Impact on revenue: none

    • Impact on EPS: +$0.11 per share (+0.6%) vs consensus FY27

    • How we got the number (estimate): a $10B buyback ÷ $361.54 ≈ 28M shares ≈ 0.56% of 4.94B → $19.39 × 0.56% ≈ +$0.11

    • Expected value: 0.45 × $0.11 ≈ +$0.05

    • What it is: a larger dividend and use of the $10.1B buyback authority before it expires on 2026-12-31 [1].

    • Why it's not priced in: only ~$0.65B was bought back in Q2–Q3 combined, as cash went to debt [1].

    • Trigger to watch: the Q4 release, and the buyback authority being renewed.

    • Why this probability: the dividend rises every December, but guarantee optics argue for keeping cash.

  • Anthropic IPO prices and de-risks the backstop — 55% chance by 2027-04-30

    • Impact on revenue: none vs consensus (consensus already includes Anthropic's 5 GW)

    • Impact on EPS: not quantifiable. The note conversion price is not public, and any equity gain would sit outside non-GAAP EPS.

    • How we got the number (estimate): n/a

    • Expected value: not quantifiable in EPS. The value is a lower chance of the biggest headwind below.

    • What it is: a public listing raising up to ~$100B at a reported ~$2T valuation [21,39].

    • Why it's not priced in: the stock still trades at a discount on the financing fears.

    • Trigger to watch: a public S-1 on EDGAR. None existed as of 2026-10-04. Bloomberg says marketing could start the week of 2026-11-09, which needs a public S-1 about 15 days earlier [39].

    • Why this probability: lowered from 62%. Six weeks later there is still no public S-1, the Pentagon dispute is unresolved and AI stocks sold off, but the banks are mandated and the Q2 profit helps.

Headwinds NOT in the forecasts

  • An AI-lab funding squeeze cuts FY28 AI revenue >20% below ~$230B — 30% chance by FY28

    • Impact on revenue: −$46B (−17%) vs ~$270B consensus FY28

    • Impact on EPS: −$4.32 per share (−14%) vs ~$30 consensus FY28

    • How we got the number (estimate): 20% × $230B = $46B × 55% × 0.84 ÷ 4.94B ≈ −$4.32

    • Expected value: 0.30 × −$4.32 ≈ −$1.30

    • What it is: Anthropic or OpenAI cannot finance 2027–28 deployments on time. Gigawatts slip and Broadcom's guarantees come into play.

    • Why it's not priced in: consensus FY28 sits close to management's path [35].

    • Trigger to watch: the Anthropic IPO slipping or shrinking; the OpenAI debt failing to close by year-end; "line of sight" wording softening on 2026-12-09.

    • Why this probability: unchanged. OpenAI's revenue restatement (10-08) is a new negative, while Anthropic's Q2 profit is a new positive.

  • Google's next flagship TPU moves mainly to MediaTek — 25% chance within 36 months

    • Impact on revenue: −$15B (−5.6%) vs ~$270B consensus FY28

    • Impact on EPS: −$1.41 per share (−4.7%) vs ~$30 consensus FY28

    • How we got the number (estimate): Google XPU revenue ~$40B in FY28 (estimate from Tan's "multi tens of billions annually") × ~35% share lost ≈ $15B × 55% × 0.84 ÷ 4.94B ≈ −$1.41

    • Expected value: 0.25 × −$1.41 ≈ −$0.35

    • What it is: MediaTek leads TPU v9.

    • Why it's not priced in: consensus assumes Broadcom stays Google's lead partner.

    • Trigger to watch: MediaTek's generation-two ASIC in early 2028; MediaTek's Q3 call (late October); TPU v9 attribution.

    • Why this probability: TrendForce's report is supply-chain hearsay, but Google clearly wants a credible second source.

  • Gross-margin dilution reaches operating margin — 35% chance in FY27

    • Impact on revenue: none

    • Impact on EPS: −$0.89 per share (−4.6%) vs $19.39 consensus FY27

    • How we got the number (estimate): FY27 operating margin 63% instead of ~66% → 3% × $173.9B × 0.84 ÷ 4.94B ≈ −$0.89

    • Expected value: 0.35 × −$0.89 ≈ −$0.31

    • What it is: HBM prices up 70–140% that cannot be fully passed through, plus MediaTek price competition.

    • Why it's not priced in: management guides one quarter at a time, and consensus gross margin was not found.

    • Trigger to watch: Q4 operating margin below 66%; the FY27 commentary.

    • Why this probability: operating margin has held so far, but HBM costs are rising faster than ever.

  • Supply-chain-risk rules force Broadcom to curb sales to Anthropic — 5% chance within 24 months

    • Impact on revenue: −$50B (−19%) vs ~$270B consensus FY28

    • Impact on EPS: −$4.70 per share (−16%) vs ~$30 consensus FY28

    • How we got the number (estimate): ~5 GW × ~$10B/GW guide-implied content ≈ $50B × 55% × 0.84 ÷ 4.94B ≈ −$4.70

    • Expected value: 0.05 × −$4.70 ≈ −$0.24

    • What it is: the label is extended to commercial suppliers.

    • Why it's not priced in: the new 10-Q risk factor is barely discussed.

    • Trigger to watch: Anthropic seeking en banc or Supreme Court review; any rule extending the label to suppliers.

    • Why this probability: as reported, the label is limited to Defense Department work, and courts are split.

  • A guarantee is called on a lab default — 8% chance within 36 months

    • Impact on revenue: none in the year of the call (beyond what the first headwind captures)

    • Impact on EPS: about −$1.01 per share one-off GAAP EPS (not non-GAAP); about −5% of FY27 consensus

    • How we got the number (estimate): a $5B net loss after rack resale ÷ 4.94B ≈ −$1.01 (tax treatment unknown)

    • Expected value: 0.08 × −$1.01 ≈ −$0.08

    • What it is: Broadcom pays out under the 85% backstop or the residual-value guarantees.

    • Why it's not priced in: fair value is booked as "not material" [1].

    • Trigger to watch: missed lease payments; Anthropic adding restricted cash.

    • Why this probability: that requires a default within three years. Both labs have just raised record rounds.

  • Software growth slows at the 2027 VMware renewals — 30% chance in FY27

    • Impact on revenue: −$2B (−1.1%) vs $173.9B consensus FY27

    • Impact on EPS: −$0.29 per share (−1.5%) vs consensus FY27

    • How we got the number (estimate): $2B × 84% segment margin × 0.84 ÷ 4.94B ≈ −$0.29

    • Expected value: 0.30 × −$0.29 ≈ −$0.09

    • What it is: underlying growth falls from ~9% to low single digits as 2024 three-year contracts renew.

    • Why it's not priced in: the upfront-licence effect flatters the headline (Bear #7).

    • Trigger to watch: ARR growth below 10%; the share of upfront licence revenue.

    • Why this probability: ARR has slowed three quarters in a row, but non-cancellable contracts limit how fast growth can fall.

  • China steers state data centres off Broadcom switches — 40% chance within 24 months

    • Impact on revenue: −$1B (−0.6%) vs consensus FY27 (estimate; Broadcom's China switch revenue not disclosed)

    • Impact on EPS: −$0.09 per share (−0.5%)

    • How we got the number (estimate): $1B × 55% × 0.84 ÷ 4.94B ≈ −$0.09

    • Expected value: 0.40 × −$0.09 ≈ −$0.04

    • What it is: SASAC informal guidance [52].

    • Why it's not priced in: no order has been issued.

    • Trigger to watch: Chinese state media, or Broadcom's Q4 China commentary.

    • Why this probability: it fits the pattern seen with Micron and Nvidia. Private hyperscalers are unaffected.

  • An OpenAI structure adds another Broadcom guarantee, and a rating agency moves to a negative outlook — 60% chance (guarantee) / 25% chance (negative outlook) within 12 months

    • Impact on revenue: none

    • Impact on EPS: not quantifiable. It is a contingent liability. Refinancing only ~$7.4B through FY28 at +25bp would cost ~$0.004 per share.

    • How we got the number (estimate): $7.4B × 0.25% ÷ 4.94B ≈ $0.004

    • Expected value: ~$0.00 in EPS. The cost would show up in the multiple and in tail risk, not in earnings.

    • What it is: a ~$30–50B OpenAI tranche with Broadcom residual-value support [25,26].

    • Why it's not priced in: earnings models ignore contingent liabilities.

    • Trigger to watch: an 8-K, the FY26 10-K (~mid-December), or agency releases.

    • Why these probabilities: the WSJ and Bloomberg both report talks targeting a year-end close. S&P has already called the first tranche credit negative without acting.

Investment Risk Score: 34/100 (moderate)

# Factor Weight Measured value Score Why
1 Liquidity & funding runway 15% Cash $23.98B + $7.5B undrawn revolver; LTM FCF $39.4B; $0 debt due FY27 25 FCF-positive with modest debt. Not lower because the ~$29B backstop exceeds cash, and FY27 purchase commitments are $52.7B.
2 Leverage 10% Net debt $37.1B ≈ 0.6x LTM adj. EBITDA ($61B, estimate); ~4% fixed coupon; contingent ~$29B disclosed, BofA max ~$370B by 2029 34 Balance-sheet leverage alone would score ~22. Undisclosed, growing guarantees add ~12.
3 Dilution (last 12 months + overhang) 10% GAAP diluted shares 4,860M → 4,887M (+0.6%); outstanding +0.7% in 9 months despite 25M bought back; SBC 6.8% of revenue; $18.2B unrecognised SBC; no converts 28 +0–3% band. SBC is heavy and withholding moved to share sales.
4 Profitability 10% LTM FCF margin ~44% (41–46% by quarter); non-GAAP operating margin 67.9% 6 Far above the 15% threshold.
5 Revenue quality & concentration 10% Top 5 end customers ~55% (was ~40%); one distributor 50%; Anthropic to be #1 XPU customer in 2027; RPO $179.2B 58 The largest single customer is not disclosed (probably 20–35%, Google). Contracted, but concentration jumped in a year and is shifting toward lab-financed buyers.
6 Execution & track record 10% Revenue beat 8/8 quarters, AI met or beat 8/8; gross-margin and rack statements contradicted (Bear #3); no material weakness 22 Hits guidance ≥80% of the time. Credibility on non-revenue claims keeps it off the bottom.
7 Competitive position & disruption 10% ~60% design-partner share (Counterpoint 2027); MediaTek 15–20% target and TPU v9 reports; Nvidia #1 DC Ethernet vendor 35 A strong niche bordering on dominant, with rising second-sourcing.
8 Cyclicality & commodity/macro sensitivity 10% AI 56% of revenue and rising; non-AI semis −23% y/y in the last downcycle; FY28 needs open credit markets for two labs 58 Cyclical. Capital-markets dependence makes it more than a normal semis cycle.
9 Regulatory, legal & geopolitical 5% ~95% of outsourced wafers from TSMC; SASAC switch survey; Section 232 phase two pending; supply-chain-risk label on future #1 customer; EU CISPE; Tesco ≥£100M 50 Material but contained.
10 Valuation & expectations 10% ~20.9x forward P/E (5-yr avg ~24x); 33.7x trailing EV/EBITDA; consensus at management's FY27 path; short interest 1.1% 40 Fair. Cheaper than its history, but expectations already embed the guide.

Overall: weighted average = 0.15×25 + 0.10×34 + 0.10×28 + 0.10×6 + 0.10×58 + 0.10×22 + 0.10×35 + 0.10×58 + 0.05×50 + 0.10×40 = 3.75 + 3.4 + 2.8 + 0.6 + 5.8 + 2.2 + 3.5 + 5.8 + 2.5 + 4.0 = 34.35 → 34.

Change vs last: −18 (52 → 34).

  • Method (almost all of it): the old six-area method put 35% on "dependency" and "execution" scored at 53–74. The ten-factor method gives four factors to balance-sheet strength and profitability, where Broadcom scores 6–34.
  • Substance, worse:
    • Concentration (top five ~40% → ~55%).
    • The guarantee pipeline (BofA's $370B framing, OpenAI talks).
    • OpenAI's revenue clarification.
    • Software ex-upfront at only ~9%.
  • Substance, better:
    • The Q4 date is confirmed (2026-12-09).
    • $0 debt is due in FY27.
    • The track record is now verified across 8 quarters from filings.
  • Net substantive change: about +1 to +2 on the old scale.

Ranking check: AVGO (34) sits between APP (28) and REAX/FTAI (50). Both AVGO and APP have net-debt-light balance sheets and ~45–66% FCF margins. AVGO scores riskier than APP because its revenue is far more concentrated and it carries contingent guarantees. It is clearly safer than FTAI and REAX, which have negative GAAP operating cash flow or floating-rate leverage.

Conclusion

  • What you are really betting on: that the two AI labs can keep raising money long enough to pay for the 2027–28 gigawatts Broadcom has pre-bought supply for, so that Broadcom's role as financier is never tested. The engineering franchise and the cash machine are the safety net, not the bet.
  • The 2–3 things that will decide the outcome:
    1. Whether the Anthropic IPO and the OpenAI debt raise close within the next ~6 months.
    2. Whether operating margin holds near 66% as HBM costs jump.
    3. Whether Broadcom keeps the lead on Google's next flagship TPU.
  • Signposts confirming the bull case:
    • Anthropic files a public S-1 and prices.
    • Q4 FY26 (2026-12-09): AI revenue ≥ $21.7B, operating margin ≥ 66%, and FY27 ~$115B held or raised.
    • The FY26 10-K gives a single, capped guarantee figure.
    • A seventh XPU customer is named.
  • Signposts that would break it:
    • The Anthropic IPO is pushed into 2027 or downsized.
    • The OpenAI financing fails to close or needs a bigger Broadcom guarantee.
    • Operating margin falls below ~62%, or FY27 purchase commitments are cut.
    • MediaTek is confirmed as lead on TPU v9.
    • A rating agency moves to a negative outlook.
  • New since the last report:
    • The track record is now verified from filings. Revenue beat guidance 8/8 quarters and AI met or beat 8/8. That is the strongest bull.
    • Software growth is mostly accounting. Excluding upfront licences ($3.47B vs $1.92B), Q3 software grew ~8.6%, not 29%.
    • The guarantee question got bigger.
      • The backstop alone (~$29B) exceeds cash.
      • BofA's model puts maximum residual-value exposure at ~$370B by 2029.
      • Management's June "no racks / not backstopped" framing was overtaken by its own 10-Q within weeks.
      • OpenAI's 10-08 revenue clarification knocked the stock −4.4%.

Open questions & data gaps

  • The total cap on guarantees. High importance. No filing sizes the residual-value support on the $42B Class A tranche or any OpenAI structure. The FY26 10-K (~mid-December) is the next chance.
  • The largest end customer's share. High importance. The 10-Q gives only top five (~55%) and the 50% distributor. A >20% single customer would move factor 5.
  • FY28 consensus. Medium-high importance. The only read (~$270B / ~$30 EPS) is from a paywalled blog preview citing Koyfin. It is the basis for three tailwind/headwind sizings.
  • Rating-agency status. Medium importance. No primary 2026 Moody's, S&P or Fitch release was found after January's A3 / A- / BBB+ positive outlooks. The S&P "credit negative" item and the Moody's "financial flexibility" quote remain secondary or UNCONFIRMED.
  • Covenants. Low importance. The credit agreement was not read. The filings only say Broadcom is "in compliance".
  • TPU 8t vs 8i attribution. Medium importance. Still unresolved: the Broadcom call vs Wccftech/UDN, with Google silent.
  • Anthropic note terms. Medium importance. Conversion price, coupon and conditions were not found. All Anthropic figures come from press reports on a confidential filing.
  • Transcripts. All call quotes are from Motley Fool and Webull machine transcripts with speaker errors. Broadcom's IR site blocks automated access.
  • Insider prices. Samueli's 2026-09-23 sale price was not found. The 10b5-1 status of Brazeal's sales is not confirmed.
Sources
  1. Broadcom Form 10-Q, quarter ended 2026-08-02 — SEC EDGAR — 2026-09-10 — https://www.sec.gov/Archives/edgar/data/1730168/000173016826000080/avgo-20260802.htm
  2. Broadcom Q3 FY26 earnings release (Ex. 99.1) — SEC EDGAR — 2026-09-02 — https://www.sec.gov/Archives/edgar/data/1730168/000173016826000076/avgo-08022026x8kxex99.htm
  3. Broadcom Form 10-K FY2025 — SEC EDGAR — 2025-12-18 — https://www.sec.gov/Archives/edgar/data/1730168/000173016825000121/avgo-20251102.htm
  4. Broadcom Q4 FY25 earnings release (Ex. 99.1) — SEC EDGAR — 2025-12-11 — https://www.sec.gov/Archives/edgar/data/1730168/000173016825000116/avgo-11022025x8kxex99.htm
  5. Broadcom Q1 FY26 earnings release (Ex. 99.1) — SEC EDGAR — 2026-03-04 — https://www.sec.gov/Archives/edgar/data/1730168/000173016826000011/avgo-02012026x8kxex99.htm
  6. Broadcom Q2 FY26 earnings release (Ex. 99.1) — SEC EDGAR — 2026-06-03 — https://www.sec.gov/Archives/edgar/data/1730168/000173016826000051/avgo-05032026x8kxex99.htm
  7. Broadcom Q4 FY24 to Q3 FY25 earnings releases (Ex. 99.1) — SEC EDGAR — 2024-12-12 to 2025-09-04 — https://www.sec.gov/Archives/edgar/data/1730168/000173016825000094/avgo-08032025x8kxex99.htm (and Q4 FY24, Q1 FY25, Q2 FY25 releases in the same EDGAR folder series)
  8. Broadcom Q3 FY26 earnings call transcript — Motley Fool — 2026-09-09 (call 2026-09-02) — https://www.fool.com/earnings/call-transcripts/2026/09/09/broadcom-avgo-q3-2026-earnings-call-transcript/
  9. Broadcom Q3 FY26 earnings call transcript (automated) — Benzinga via Webull — 2026-09-02 — https://www.webull.com/news/15511095937917952
  10. Broadcom Q2 FY26 earnings call transcript — Motley Fool — 2026-06-03 — https://fool.com/earnings/call-transcripts/2026/06/03/broadcom-avgo-q2-2026-earnings-transcript/
  11. Broadcom Q1 FY26 earnings call transcript — Motley Fool — 2026-03-04 — https://www.fool.com/earnings/call-transcripts/2026/03/04/broadcom-avgo-q1-2026-earnings-call-transcript/
  12. Broadcom Q4 FY25 earnings call transcript — Motley Fool — 2025-12-12 — https://www.fool.com/earnings/call-transcripts/2025/12/12/broadcom-avgo-q4-2025-earnings-call-transcript/
  13. Broadcom 8-K, Google long-term agreement and Anthropic ~3.5 GW — SEC EDGAR — 2026-04-06 — https://www.sec.gov/Archives/edgar/data/1730168/000119312526144028/d87999d8k.htm
  14. Broadcom 8-K, Apple agreements through 2031 — SEC EDGAR — 2026-07-06 — https://www.sec.gov/Archives/edgar/data/1730168/000119312526295589/d84378d8k.htm
  15. Broadcom free writing prospectus (ratings A3/A-/BBB+, positive outlooks) — SEC EDGAR — 2026-01-07 — https://www.sec.gov/Archives/edgar/data/1730168/000119312526005084/d842461dfwp.htm
  16. Broadcom DEF 14A proxy statement — SEC EDGAR — 2026-03-02 — https://www.sec.gov/Archives/edgar/data/1730168/000119312526085691/d49254ddef14a.htm
  17. Broadcom EDGAR submissions index — SEC — retrieved 2026-10-12 — https://data.sec.gov/submissions/CIK0001730168.json
  18. Broadcom insider transactions (issuer table) — SEC EDGAR — retrieved 2026-10-12 — https://www.sec.gov/cgi-bin/own-disp?action=getissuer&CIK=1730168
  19. AVGO insider transactions — Quantisnow — retrieved 2026-10 — https://www.quantisnow.com/insiders/AVGO
  20. Broadcom to lend Anthropic up to $42 billion to lease its chips, filing says — Reuters via CNBC — 2026-10-01 — https://www.cnbc.com/2026/10/01/broadcom-lending-anthropic-42-billion-chips-reuters.html
  21. Anthropic IPO S-1 prospectus income statement — Fortune — 2026-09-29 — https://fortune.com/2026/09/29/anthropic-ipo-s-1-prospectus-income-statement/
  22. Anthropic's IPO filing shows 12-fold revenue jump, $518B compute bill — InvestmentNews — 2026-09-29 — https://investmentnews.com/equities/anthropics-landmark-ipo-filing-shows-12-fold-revenue-jump-518b-compute-bill/268391
  23. Broadcom is Anthropic's chip supplier, lender and debt guarantor (relaying Bloomberg) — TechTimes — 2026-10-06 — https://www.techtimes.com/articles/328625/20261006/broadcom-anthropics-chip-supplier-lender-debt-guarantor-60-billion-syndication-begins.htm
  24. Chips with everything: Broadcom backstops US$42bn funding for Anthropic — IFR — 2026-10 — https://www.ifre.com/loans/2496132/chips-with-everything-broadcom-backstops-us42bn-funding-for-anthropic
  25. Broadcom plans $50 billion financing for OpenAI chips, WSJ says — Bloomberg — 2026-10-07 — https://www.bloomberg.com/news/articles/2026-10-07/broadcom-plans-50-billion-financing-for-openai-chips-wsj-says
  26. Broadcom eyes US$30bil AI chip financing deal (Bloomberg) — The Star — 2026-10-09 — https://www.thestar.com.my/business/business-news/2026/10/09/broadcom-eyes-us30bil-ai-chip-financing-deal
  27. Why Broadcom stock took a dive today — Motley Fool — 2026-10-08 — https://www.fool.com/investing/2026/10/08/why-broadcom-stock-took-a-dive-today/
  28. Bank of America downgrades Broadcom credit to Neutral (Wallstreetcn relay) — Longport — 2026-08-10 — https://longportapp.cn/news/295504198
  29. Broadcom's AI financing could reach $370 billion — Nasdaq — 2026-08 — https://nasdaq.com/articles/broadcoms-ai-financing-could-reach-370-billion-its-not-bad-it-sounds ; Broadcom's $60 billion AI debt deal hides a $370 billion question — 24/7 Wall St — 2026-08-24 — https://247wallst.com/investing/2026/08/24/broadcoms-60-billion-ai-debt-deal-hides-a-370-billion-question-nobody-on-wall-street-wants-to-answer/
  30. S&P calls Broadcom's first AI tranche credit negative — CryptoBriefing — 2026-06-11 — https://cryptobriefing.com/sp-broadcom-ai-tranche-credit-negative/ ; Broadcom backs $35bln AI chip order (The Information relay) — Newsquawk — 2026-06-15 — https://www.newsquawk.com/headlines/broadcom-avgo-reportedly-backs-a-usd-35bln-ai-chip-order-to-boost-demand-with-plans-to-finance-over-20-gigawatts-of-compute-capacity-potentially-impacting-credit-through-chip-lease-financial-backstop-reports-the-information
  31. Broadcom CDS surges to record 122 basis points — CryptoBriefing — 2026-08-21 — https://cryptobriefing.com/broadcom-cds-record-debt-deal/
  32. Broadcom credit risk soars on mega AI debt financing backstops — Bloomberg Government — 2026-08-24 — https://news.bgov.com/legal-ops-and-tech/broadcom-credit-risk-soars-on-mega-ai-debt-financing-backstops
  33. AVGO analyst forecast and estimates — StockAnalysis — data to 2026-10-09 — https://stockanalysis.com/stocks/avgo/forecast/
  34. AVGO statistics and financial ratios history — StockAnalysis — 2026-10-09 — https://stockanalysis.com/stocks/avgo/statistics/ ; https://stockanalysis.com/stocks/avgo/financials/ratios/
  35. Spare a thought for Broadcom shareholders (FY28 consensus via Koyfin chart; paywalled preview) — MBI Deep Dives — 2026-09-04 — https://www.mbi-deepdives.com/avgo3q26/
  36. These analysts increase their forecasts on Broadcom following upbeat Q3 earnings — Benzinga (search snippet; page 403) — 2026-09 — https://www.benzinga.com/analyst-stock-ratings/price-target/26/09/61600199/these-analysts-increase-their-forecasts-on-broadcom-following-upbeat-q3-earnings
  37. Macquarie rating history — StreetInsider / Finbold — 2026-06-04 and 2026-09-02 — https://www.streetinsider.com/rating_history.php?firm=38 ; https://finbold.com/analysts-set-broadcom-stock-price-target-2/
  38. Broadcom sees $230B AI revenue (Moorhead commentary) — Yahoo Finance — 2026-09 — https://finance.yahoo.com/technology/ai/articles/broadcom-sees-230b-ai-revenue-023424926.html
  39. Anthropic's public S-1 still missing from EDGAR — Yahoo Finance — 2026-10-04 — https://finance.yahoo.com/markets/stocks/articles/anthropic-public-1-still-missing-210349323.html
  40. Our eighth generation TPUs — Google Blog — 2026-04-22 — https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/eighth-generation-tpu-agentic-era/
  41. Google splits TPUv8 strategy: Broadcom training, MediaTek inference — Wccftech — 2026-04-20 — https://wccftech.com/google-splits-tpuv8-strategy-two-chips-broadcom-training-mediatek-inference-duties/amp/
  42. MediaTek earnings call coverage — TrendForce — 2026-08-03 — https://www.trendforce.com/news/?p=61938
  43. MediaTek reportedly wins TPU v9 orders on SerDes shift — TrendForce — 2026-06-23 — https://www.trendforce.com/news/2026/06/23/news-mediatek-seen-raising-prices-10-20-reportedly-wins-tpu-v9-orders-on-serdes-shift/
  44. Marvell, AMD reportedly shake up Google TPU race — TrendForce — 2026-08-20 — https://www.trendforce.com/news/2026/08/20/news-marvell-amd-reportedly-shake-up-google-tpu-race-putting-broadcom-mediatek-under-pressure/
  45. Marvell 8-K, Google agreement and warrant — SEC EDGAR — 2026-08 — https://www.sec.gov/Archives/edgar/data/1835632/000119312526356217/d412696d8k.htm
  46. Marvell Investor Day 2026 targets $80B FY2031 revenue — Futurum — 2026-10-06 — https://futurumgroup.com/?p=96506
  47. Broadcom set to dominate custom AI chip market with 60% share by 2027 (Counterpoint) — Finviz/Benzinga — 2026-01 — https://finviz.com/news/288744/broadcom-set-to-dominate-custom-ai-chip-market-with-60-share-by-2027-counterpoint-says
  48. Broadcom redefines AI infrastructure at 2026 OCP Global Summit — Broadcom via Finviz — 2026-10-08 — https://finviz.com/news/400527/broadcom-redefines-ai-infrastructure-with-industry-leading-networking-innovations-at-2026-ocp-global-summit
  49. NVIDIA becomes #1 in datacenter Ethernet switching as 1Q26 market surges — IDC — 2026-06-17 — https://www.idc.com/resource-center/blog/nvidia-becomes-1-in-datacenter-ethernet-switching-as-1q26-market-surges-39-8-to-15-4-billion/
  50. Broadcom rides rocketing trend for custom AI accelerators — The Next Platform — 2026-09-10 — https://www.nextplatform.com/connect/2026/09/10/broadcom-rides-rocketing-trend-for-custom-ai-accelerators/5295681
  51. DC Circuit panel upholds Pentagon's ban on Anthropic, so what comes next — Breaking Defense — 2026-09 — https://breakingdefense.com/2026/09/dc-circuit-panel-upholds-pentagons-ban-on-anthropic-so-what-comes-next/
  52. China surveys Broadcom switch use in state data centers, FT reports — Reuters via The Standard — 2026-09-23 — https://www.thestandard.com.hk/innovation/article/343583/China-surveys-Broadcom-switch-use-in-state-data-centers-FT-reports
  53. Presidential action: semiconductors and critical minerals under Section 232 — Perkins Coie — 2026-01 — https://perkinscoie.com/insights/article/presidential-action-semiconductors-and-critical-minerals-under-section-232
  54. Lutnick backs broad semiconductor tariffs — roic.ai — 2026-09-02 — https://www.roic.ai/news/lutnick-backs-broad-semiconductor-tariffs-weighing-duties-on-chips-and-devices-09-02-2026
  55. CISPE competition complaint — CISPE — 2026-03-19 — https://www.cispe.cloud/?p=25724 ; Five cloud business groups urge EU interim measures against Broadcom — Reuters via The Star — 2026-07-15 — https://www.thestar.com.my/tech/tech-news/2026/07/15/five-cloud-business-groups-urge-eu-interim-measures-against-broadcom
  56. Why Tesco's taking Broadcom to court over VMware deals — Cyber Magazine — 2026 — https://cybermagazine.com/news/why-tescos-taking-broadcom-to-court-over-vmware-deals
  57. Critical VMware vCenter RCE flaw exploited for reverse SSH access — BleepingComputer — 2026-08 — https://bleepingcomputer.com/news/security/critical-vmware-vcenter-rce-flaw-exploited-for-reverse-ssh-access
  58. TrendForce 2027 HBM price forecast — TrendForce — 2026-09-29 — https://www.trendforce.com/presscenter/news/20260929-13255.html
  59. Amazon Q2 2026 earnings (capex ~$220B) — Futurum — 2026-07 — https://futurumgroup.com/?p=93607
  60. Broadcom Inc. annual revenue series (XBRL) — SEC — retrieved 2026-10-12 — https://data.sec.gov/api/xbrl/companyconcept/CIK0001730168/us-gaap/RevenueFromContractWithCustomerExcludingAssessedTax.json
  61. US and China to extend trade-war truce — Supply Chain Dive — 2026-09 — https://www.supplychaindive.com/news/us-china-to-extend-trade-war-truce-by-2-months/831251/
  62. OpenAI closes $122bn funding round — Yahoo Finance — 2026-03 — https://finance.yahoo.com/markets/stocks/articles/openai-closes-122bn-funding-round-093126735.html

Notes:

  • Stock prices, consensus and valuation multiples come from StockAnalysis (S&P Global data). Analyst targets are as publicly reported; the full reports are paywalled.
  • "Estimate" marks our own arithmetic.
  • Scores are model judgements, not facts. This is research, not advice.
All AVGO Bulls vs Bears reports
Disclaimer. AI-generated research and podcast (Claude by Anthropic, voices by Google Gemini) for educational purposes — not financial advice and not a recommendation to buy or sell. Willem (MR.INVEST) owns this stock. Check the company’s own filings before making any decision.
← All Bulls vs Bears reports
AVGO Bulls vs Bears — Stock Analysis — MR.INVEST