Stock Analysis · Bulls vs Bears

SharonAI Holdings SHAZ

🐂 Bulls vs Bears 🐻Report #213 Oct 2026
★ Willem owns this stock · 1.6% of the portfolio
73
high risk▲ +2 vs last
🐂 8vs12 🐻
Bears lead
30%
▲ Chance of the biggest tailwind the forecasts miss
60%
▼ Chance of the biggest headwind the forecasts miss
The verdict

The contracts are real and some are priced well: $4 per GPU-hour on the newest B300 deal against an implied ~$2.33 NVIDIA floor. Our re-test, though, found two new structural problems. Every phase of the $4.88B NVIDIA contract ends on the same date, so each month of build delay is revenue lost for good. And the ~$3.9B still to be funded must now be raised in a market where Firmus, its closest Australian peer, has…

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 SHAZ · AI can make mistakes — numbers in brackets point to the sources · Not financial advice

🐂 Bull arguments 8

  1. 1An ~$8.8B take-or-pay book backed by deposits and letters of creditMedium
  2. 2New contracts are priced well above the NVIDIA floorMedium
  3. 3$1.86B cash and a Goldman-led loan fund the first waveMedium
  4. 4NVIDIA's six-year minimum makes the 72 MW build financeableMedium
  5. 5The market values the company below its 2027 consensus revenueMedium
  6. 6The largest holder paid ~$65 a share and has not soldMedium
  7. 7Secured Australian power gained value after Firmus's IPO failedMedium
  8. 8A new chair, CFO, COO and Big Four auditorMedium

🐻 Bear arguments 12

  1. 1About $3.9B of commitments is still unfundedHigh
  2. 2Delays permanently shrink the NVIDIA contract and othersHigh
  3. 3The $470M run-rate promise was quietly droppedMedium
  4. 4ESDS owes ~4.7× its revenue yearly and serves a sanctioned bankMedium
  5. 5NVIDIA is supplier, floor-giver and judge of uptimeHigh
  6. 6The first loan is expensive, short and parent-guaranteedMedium
  7. 7Reporting errors and shifting targets undermine trust in the numbersMedium
  8. 8Registered resale shares exceed the shares outstandingMedium
  9. 9The neocloud funding window is closingMedium
  10. 10Unproven at scale: ~4 MW live, no security certificationsMedium
  11. 11Founder control and a CEO with unresolved Mawson disputesLow
  12. 12Blackwell rental prices are near their peakLow

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 #1An ~$8.8B take-or-pay book backed by deposits and letters of credit

  • The point: SharonAI pre-sold its capacity before buying the chips. If it delivers, revenue goes from ~$2M a quarter to over $1B a year, on contracts customers pay whether or not they use them.
  • Evidence:
    • TCV (the company's "total contract value" metric) is "over US$8.8bn" (10-01) [8].
    • The 10-Q itemises the five big contracts [3,11]:
      • NVIDIA: up to $4.88B, six years;
      • AI lab: $1.32B, five years, New Zealand;
      • ESDS: ~$1.26B;
      • APAC technology company: ~$950M;
      • AI platform: $373M.
    • The 10-Q calls three of them five-year take-or-pay [11].
    • Customer deposits were $143.9M at 2026-06-30, against zero at 2025-12-31 [3].
    • The APAC customer accepted the first cluster on 2026-08-20, which "triggers release of cash security currently held in escrow" [21].
  • Why it matters:
    • Interpretation: if the contracts start on the dates given, they alone add up to roughly $1.2–1.3B of 2027 revenue (estimate; see Bear #3). That is close to the ~$1.54B consensus.
    • Lenders lend against such contracts, as the first facility shows [7].
  • 😈 Devil's advocate:
    • TCV is "an operating metric and does not represent revenue recognized" [11]. Remaining performance obligations (RPO, the audited version of a backlog) are not disclosed in any filing [3,5].
    • Over half the book is NVIDIA's falling floor (Bear #5).
    • The second-biggest named counterparty, ESDS, is small relative to what it owes (Bear #4).
    • Three of the five counterparties are unnamed.
  • Verdict: Medium The contracts are real and the first acceptance happened. Their value depends on on-time delivery and on two weak or opaque counterparties.

🐂 Bull #2New contracts are priced well above the NVIDIA floor

  • The point: Demand for the newest NVIDIA chips is strong enough that SharonAI signs new deals well above the price NVIDIA guarantees on its 72 MW. That makes reselling NVIDIA capacity above the floor plausible.
  • Evidence:
    • Q2 call, Manning on the $373M AI-platform deal: "a B300 deployment with a record price of over $4 per GPU-hour" [14]. This confirms a figure that our report #1 had only from a summary.
    • The NVIDIA floor implied by $4.88B over six years and 40,000 GPUs is about $2.33 per GPU-hour on average (arithmetic: $4.88B ÷ (40,000 × 6 × 8,760 h)). Compass Point computed ~$2.32 [14,40].
    • Silicon Data's B200 rental index was $5.62 per GPU-hour on 2026-09-01, up 27.6% this year, with August volatility the lowest of 2026 [33].
    • NVIDIA's own 10-Q says AI clouds "can unilaterally stop providing [capacity] to us and sell to third-party customers at more advantageous rates" [24].
  • Why it matters: each GPU-hour resold at $4 instead of the ~$2.33 floor adds ~$1.67 of revenue before the revenue share. On 40,000 GPUs that is up to ~$585M a year of gross spread (arithmetic, upper bound).
  • 😈 Devil's advocate:
    • The revenue-share percentage is not disclosed. SharonAI's 10-Q says investors "should not rely" on it [3].
    • The NVIDIA capacity only arrives April–August 2027 [3]. Silicon Data's 12-month forward B200 price is ~12% below spot, so the market already expects lower prices then [33].
    • The >$4 deal is 2,048 GPUs. One deal is not a price trend.
  • Verdict: Medium A real data point, but it is one contract and the upside only starts in 2027.

🐂 Bull #3$1.86B cash and a Goldman-led loan fund the first wave

  • The point: It raised enough to deliver the ESDS, APAC, AI-lab and AI-platform contracts, and proved that lenders will finance its contracts.
  • Evidence:
    • Cash was $1.86B at 2026-06-30 [3]. H1 financing brought in $586.9M of common stock, $438.1M of pre-funded warrants and $1,065.6M of convertibles [3].
    • The $356M facility (signed 2026-09-28) was arranged by Jarden, with "Goldman Sachs and select large private credit funds" as lenders [7,8].
    • It becomes non-recourse to the parent once Facility A is repaid from an escrowed customer deposit after acceptance [7].
    • The company calls it "the first in an expected series of GPU financings" [8].
    • Q1 call, Manning: "we're funded fundamentally from all the equity components for all the currently announced contracts" [13]. That was said before the NVIDIA deal.
  • Why it matters: if each contract can carry its own project debt, the remaining build can be funded with less new equity.
  • 😈 Devil's advocate:
    • Commitments are ~$6.1B against ~$2.2B of cash plus facility. The NVIDIA hardware alone is ~$4B [3] (Bear #1).
    • $356M against ESDS/APAC capex of ~$733M is a loan-to-value of only ~49%, below the 70–80% the company had targeted [5]. That points to weaker borrowing capacity per dollar of contract.
    • Drawn amount: not found. The availability period ended about 2026-10-28 [7].
  • Verdict: Medium It has the money for the non-NVIDIA contracts, not the whole plan.

🐂 Bull #4NVIDIA's six-year minimum makes the 72 MW build financeable

  • The point: A floor from NVIDIA, a AAA-type credit, is the collateral that lets a small company borrow billions.
  • Evidence:
    • The NVIDIA contract is "up to $4.88 billion", six years, with "revenue share, reporting and audit mechanics" [10].
    • Q2 call: "$4.9 billion of minimum revenue", an "average of $817 million of revenue per annum", and "pricing… is guaranteed as a minimum only" [14].
    • NVIDIA's 10-Q confirms the program: its AI-cloud commitments are "typically six years in duration" and total $36B at 2026-07-26. NVIDIA also guarantees land, power and shell leases for AI clouds, with a maximum gross exposure of $3.5B [24].
    • No "pause" language appears in the 10-Q. NVIDIA told the WSJ the model "is still in place and continues to evolve due to high demand" [24,25].
  • Why it matters: ~$817M a year from one counterparty should support project debt at a lower rate than the first facility.
  • 😈 Devil's advocate:
    • NVIDIA can terminate for material breach, including "frequent instances of GPU cluster unavailability" [10]. SharonAI has never run a cluster this size.
    • The program was paused for new deals in August over antitrust concerns [25]. SharonAI still has not said whether its contract is affected.
    • Firmus, the other launch partner, could not get its IPO done (Bear #9). That suggests the market does not treat the NVIDIA backstop as decisive.
  • Verdict: Medium The floor is contractual and NVIDIA confirms the program exists. Its value depends on uptime SharonAI has not yet shown.

🐂 Bull #5The market values the company below its 2027 consensus revenue

  • The point: At $35.69 the market cap is about $1.37B (38.48M shares). That is below 2027 consensus revenue of ~$1.54B and below the 2026-06-30 cash of $1.86B.
  • Evidence:
    • Close $35.69 on 2026-10-12. The stock was $63.20 on 09-22 and $68.73 in the June placement [29,43].
    • Shares: 38,341,987 Class A (13D/A denominator, 2026-09-17) plus 136,341 Class B [15,5].
    • Consensus: FY2026 revenue $118.6M (5 analysts, range $88–152M); FY2027 ~$1.54B. All 6 analysts rate it Buy, with targets $80–124 (average $105) [29,40].
    • Market cap ÷ FY2027 consensus revenue ≈ 0.89×. CoreWeave trades at 2.3× forward sales, IREN at 4.9× and Nebius at 8.0× [29].
  • Why it matters: if even the contracted part of 2027 revenue arrives, the current price leaves room for some dilution.
  • 😈 Devil's advocate:
    • The cash is not spare: it is earmarked for ~$6.1B of commitments [3].
    • The low multiple is the market pricing in the dilution needed to fund the NVIDIA build (Headwind #1). Per share, the 2027 revenue belongs partly to shareholders who have not bought yet.
    • The consensus revenue assumptions (MW live, $/MW) are not public, and the FY2028 consensus was not found.
  • Verdict: Medium Weak. Cheap on paper, but the share count it rests on is not final.

🐂 Bull #6The largest holder paid ~$65 a share and has not sold

  • The point: Situational Awareness LP, Leopold Aschenbrenner's AI-focused fund, owns 21.1% and is ~45% under water, yet shows no sales.
  • Evidence:
    • 8,070,950 shares bought for $523.9M (~$64.9 each). It exercised its pre-funded warrants on 06-30 and 09-15 [15].
    • Since the IPO there have been only six Form 4s: grants, one share issue in lieu of expenses, and warrant exercises. There were no open-market sales and no 10b5-1 plans [15].
    • Millennium doubled its position from 870,771 to 1,742,894 shares (13G/A, 08-14) [17].
  • Why it matters: a committed anchor holder lowers the risk of a disorderly exit and could back a later round.
  • 😈 Devil's advocate:
    • Oaktree cut its position from 1,000,000 shares (6.25%) to 857,053 (4.99%, of which 567,507 are convert shares) by 08-10 [16]. Interpretation: the convert anchor sold common stock.
    • Situational Awareness reportedly sold much of its other public book in July (previous report, secondary).
    • Founders are locked up only until 2027-03-31 [18].
  • Verdict: Medium Weak. A positive signal, but no insider buying and one anchor already trimmed.

🐂 Bull #7Secured Australian power gained value after Firmus's IPO failed

  • The point: SharonAI has 212 MW "secured" (120 MW contracted, 92 MW available) in a power-constrained country. Its main local rival just lost its public funding route.
  • Evidence:
    • 212 MW secured, up 80 MW on Q2; "early megawatts potentially as early as late this year, but definitely in Q1 next year" [14].
    • Firmus withdrew its ~US$5B ASX IPO on 2026-10-08/09 after ~48 hours of weak demand, at a planned US$30.6B valuation [31,42].
    • NEXTDC's FY26 results show 565 MW of forward orders, with 197 MW converting to billing in FY27. Its build programme is "on schedule" [37].
  • Why it matters: if Firmus slows down, there is less competing Australian sovereign-AI supply for the 92 MW SharonAI has not yet sold.
  • 😈 Devil's advocate:
    • Firmus still has US$2.85B of fresh private money from NVIDIA, Coatue, Blackstone and Jane Street, and says it will seek private capital [42].
    • SharonAI owns neither the power nor the buildings. NEXTDC sells to hyperscalers too, and does not name SharonAI in its FY26 documents [37].
    • The same weak demand that sank Firmus's IPO hits SharonAI's own funding (Bear #9).
  • Verdict: Medium Weak. Scarce power is a real asset, but the Firmus read-across cuts both ways.

🐂 Bull #8A new chair, CFO, COO and Big Four auditor

  • The point: The founder-run company is being professionalised, which large customers and lenders require.
  • Evidence:
    • Chair Andrew Penn (ex-Telstra CEO). CFO Anuj Goel (ex-Macquarie Capital) since 2026-08-24. COO David Burns (ex-Telstra) since 2026-09-07 [22,23].
    • EY was engaged on 2026-09-23 after a "competitive process involving several leading firms". HoganTaylor's letter agrees with the company's statements, with no disagreements [9].
    • Founder consulting agreements were terminated in April 2026 "with no notice period or payment" [3].
  • Why it matters: a credible CFO and auditor should lower the cost of the next facility and help remediate the material weakness.
  • 😈 Devil's advocate:
    • HoganTaylor was dismissed on 2026-09-20, 24 days after shareholders ratified it on 08-27 [9]. EY is the third auditor in nine months.
    • The material weakness over "complex financial instruments" was still open at 06-30 [3].
    • Co-founder Leece was moved out of the COO role into a fixed-term job ending 2027-03-31, the same day the founder lock-up ends [18,22].
  • Verdict: Medium Weak. Right direction, but not yet tested by an EY-reviewed quarter.

Bear case

🐻 Bear #1About $3.9B of commitments is still unfunded

  • The point: SharonAI has signed to spend about three times the money it has.
  • Evidence:
    • Purchase and infrastructure commitments were "approximately $6 billion" at 06-30 [3]. Itemised [3]:
      • NVIDIA hardware ~$4B;
      • ESDS ~$720M;
      • AI lab $765.1M;
      • APAC ~$400M;
      • AI platform $217.2M.
    • Cash of $1.86B plus the $356M facility is ~$2.2B, leaving a gap of ~$3.9B (arithmetic, before deposits, prepayments and contract cash flows).
    • No year-by-year schedule is disclosed. The NEXTDC phases (April–August 2027) and the "64,000+ GPUs by mid-2027" target put most spending between Q4 2026 and mid-2027 [3].
    • Q2 burn: operating cash flow excluding the $143.9M deposit was about −$113M, plus $268M of equipment advances (arithmetic from the 10-Q) [3].
    • The 6% notes purchase agreement bars new equity for a year after its effective date, except a placement at ≥$55 or ≤$25M to a strategic partner. The exact end date was not found [18]. Interpretation: a plain stock sale at today's ~$36 may need a waiver, a convertible or a preferred.
  • Why it matters: the gap will be filled by expensive debt, new shares, or by giving up part of the NVIDIA build. Each option takes value from today's shareholders.
  • 🛡️ Counter-argument: NVIDIA's floor should support project debt at 70–80% loan-to-value. Prepayments of 10–20% fund part of the hardware. NVIDIA also guarantees land, power and shell leases for AI clouds [24].
  • Severity: High It is the single biggest issue, and no financing for the NVIDIA build has been announced in the four months since signing.

🐻 Bear #2Delays permanently shrink the NVIDIA contract and others

  • The point: Being late doesn't just postpone revenue. It cancels it.
  • Evidence:
    • NVIDIA 8-K: phases are delivered and accepted separately, but "all phases are required to terminate on the same date regardless of when delivered and accepted" [10]. This is new in this report.
    • Other contracts: "for each day of delay beyond the handover date, the total service period may be reduced", and customers may terminate if acceptance tests miss deadlines [3].
    • Track record: the 10-K guided B200/B300 clusters for "1H 2026"; the first acceptance came on 2026-08-20 [1,21].
    • On the Q2 call, Manning would not confirm the 09-16 ESDS start date put to him by B. Riley ("We've got RFS dates… end of Q3 turning on"). He named hardware delivery and data-centre readiness as the main risks [14].
    • GreenSquareDC SYD1 was targeted ready-for-service on 2026-09-26. No confirmation was found in any filing [3].
  • Why it matters: each month the average NVIDIA phase slips costs ~$68M of floor revenue for good (arithmetic: $817M ÷ 12). Rent, interest and depreciation keep running meanwhile.
  • 🛡️ Counter-argument: SharonAI leases finished halls rather than building them. NEXTDC says its expansions are "on schedule" [37]. Lenovo, Cisco and WWT do much of the integration.
  • Severity: High The contracts make delay costs irreversible, and the company has already slipped once.

🐻 Bear #3The $470M run-rate promise was quietly dropped

  • The point: Management set a public revenue target, then stopped repeating it without explanation.
  • Evidence:
    • Q1 call (2026-05-15), Manning: "we expect our exit 2026 revenue run rate of at least $470 million" [13]. Report #1 had this from one summary only; it is now confirmed.
    • Q2 call (2026-08-06): no run-rate figure. Instead: "we expect the first material revenue to commence in the fourth quarter of 2026" and "a full month and a bit of billing in Q3 for the B300" [14].
    • Consensus FY2026 revenue of $118.6M needs ~$116M in H2, after $2.2M in H1 [3,29].
    • Estimate: ESDS at full rate is ~$63M a quarter ($1.26B ÷ 20 quarters) and APAC ~$47.5M a quarter ($950M ÷ 20). Both fully live for all of Q4 gives ~$110M. So consensus needs essentially everything running on time.
    • Estimate for FY2027 contracted revenue on schedule:
      • NVIDIA ~$475M (7 of 12 months at $817M a year);
      • ESDS ~$252M;
      • APAC ~$190M;
      • AI lab ~$200M (about 9 months);
      • AI platform ~$67M;
      • total ~$1.18B.
    • Against consensus of ~$1.54B, that leaves ~$350M expected from unsold capacity or above-floor resale. The NVIDIA floor is front-loaded (it "declines predictably"), so the gap may be smaller, ~$250M.
  • Why it matters: both years' consensus needs perfect timing plus new sales. Q3 results (date not announced; estimated early to mid-November) are the first test.
  • 🛡️ Counter-argument: The first APAC cluster was accepted in August [21]. Q4 billing plus the new 80 MW could close the gap. Management may simply have stopped giving targets on lawyers' advice after the IPO.
  • Severity: Medium High. A miss would hit the stock's credibility just when it needs to raise money.

🐻 Bear #4ESDS owes ~4.7× its revenue yearly and serves a sanctioned bank

  • The point: The ~$1.26B ESDS contract is ~$252M a year from a company with ~$53M of annual revenue, whose largest client was a Russian bank on the US sanctions list.
  • Evidence:
    • ESDS FY26 revenue ₹472.2 crore ($53M) and profit ₹120.8 crore ($14M) (Indian press, from the IPO prospectus) [27].
    • Its Aug–Sep 2026 IPO raised ₹720 crore fresh (~$81M), of which ₹576 crore goes to its own equipment [27].
    • ESDS's draft prospectus (primary): the top client for the six months to 2024-09-30 was "a BFSI incorporated in Russia" on the OFAC SDN list (the US Treasury's sanctions list). It made up 22.02% of revenue (₹379.09M) and paid in rupees through a Special Rupee Vostro Account [26].
    • The prospectus does not name the bank. Bleecker Street (short seller) says it is Gazprombank, which is on the SDN list [28,41]. The name is unconfirmed; the fact of a sanctioned Russian bank is confirmed.
    • SharonAI's own S-1/A: it is "exposed to the credit quality of ESDS" [5].
    • The $140M of letters of credit or guarantees from ESDS is in report #1's reading of the 10-Q, but helper re-reads did not find it in the latest filings. Treat it as not re-confirmed.
  • Why it matters: the $356M facility carries a "review event" if a customer contract default gives a termination right [7]. A failed counterparty also leaves ~$720M of hardware to re-let.
  • 🛡️ Counter-argument: ESDS's IPO was 136× subscribed and its shares rose ~200%, so it can raise more equity. Interpretation: it is probably a reseller passing the cost through to its own clients. ESDS's latest top client was 15.93% of FY26 revenue, so the Russian share may have fallen [27].
  • Severity: Medium High. Primary documents now confirm the sanctions link that report #1 could not verify.

🐻 Bear #5NVIDIA is supplier, floor-giver and judge of uptime

  • The point: ~55% of TCV depends on one counterparty that sells SharonAI its chips, sets a falling price, and can walk away if clusters are down too often.
  • Evidence:
    • Termination is allowed for material breach including "frequent instances of GPU cluster unavailability". Cure is allowed only for "certain material breaches (not all)", and "NVIDIA may not accept a cure" [10].
    • The price "declines predictably over the six-year contract term" [3].
    • The floor averages ~$2.33 per GPU-hour, against $4+ in the market (Bull #2).
    • The WSJ reported on 2026-08-27/28 that NVIDIA paused the program for new deals. SharonAI fell ~8% pre-market and has not commented since [25].
  • Why it matters: if SharonAI breaches on uptime in its first large deployment, it could own ~$4B of GPUs with only market prices to rely on.
  • 🛡️ Counter-argument: NVIDIA's 10-Q still lists $36B of these commitments with no pause language [24]. Dropping a named launch partner would damage NVIDIA with every neocloud. Lucid's analyst said NVIDIA continues to work with SharonAI (secondary) [40].
  • Severity: High The uptime termination clause is new to this analysis and turns execution risk into contract risk.

🐻 Bear #6The first loan is expensive, short and parent-guaranteed

  • The point: SharonAI pays roughly 3 percentage points more than neoclouds with strong customers, and must refinance before its contracts end.
  • Evidence:
    • Facility A: 7.25% stepping to 9.95%. Facility B: 9.95% fixed, "excluding fees". Default interest +2% [7,8].
    • It is a single lump-sum repayment 42 months after first drawdown, while the customer contracts run five years [7].
    • The parent guarantees both tranches until Facility A is repaid [7].
    • "Review events": change of control, a customer-contract default, manager events, or operating costs above a redacted share of the model. Each opens a 30–45 day window that ends in acceleration rights [7].
    • Services falling below a redacted level for 3 consecutive days, or 6 days in a month, is an immediate default [7].
    • Peers:
      • Lambda: $1.0B at 6.78% fixed, amortising, investment-grade rated (10-01) [34];
      • Nebius: $775M at SOFR+2.50% [35];
      • CoreWeave unsecured notes: 9.625–9.75% [36].
    • Fixed cash interest is ~$85.7M a year if fully drawn (arithmetic: $21M on the 6% notes + $33.3M on the 4.75% notes + ~$31.4M on the facility).
  • Why it matters: at ~10%, debt eats much of the margin on a GPU whose rental price falls each year. The parent guarantee means the "non-recourse" label isn't true until after acceptance.
  • 🛡️ Counter-argument: it is a first-time borrower's price, and it should fall once a cluster has a track record. Facility A becomes non-recourse soon after acceptance.
  • Severity: Medium

🐻 Bear #7Reporting errors and shifting targets undermine trust in the numbers

  • The point: Investors have to rely on management's metrics (TCV, MW, GPUs), and those metrics and the accounts have repeatedly changed.
  • Evidence:
    • The 10-Q/A (08-11) corrected 2025 EPS and share counts: Q2-25 EPS went from −$2.41 to −$0.27 [4].
    • The Q2 press release shows Q2-26 EPS of −$26.16 on 16.4M weighted shares. The 10-Q shows −$22.84 on 18.8M. The two disagree and were never reconciled [3,11].
    • An 8-K of 09-23 was labelled Item 4.01 (auditor change) on EDGAR but contained only a VAST Data press release [9].
    • Targets moved:
      • GPUs by mid-2027: 62,000 → 64,000 → "over 68,000" in 11 weeks;
      • secured capacity: 132 MW in the 10-Q became 212 MW in the press release two days later [3,8,11].
    • The 10-K said NVIDIA was a strategic shareholder; this was retracted on 04-13 [1].
    • A material weakness over complex financial instruments is open, and there have been three auditors in nine months [3,9].
  • Why it matters: lenders price trust. An error in the first EY-reviewed quarter would hit refinancing just when it matters.
  • 🛡️ Counter-argument: the corrections were made by the company itself and affected prior-year comparisons, not 2026 cash. EY and a Macquarie-trained CFO are the fix.
  • Severity: Medium

🐻 Bear #8Registered resale shares exceed the shares outstanding

  • The point: Much of the June money can now be sold, and the buyers paid far more than today's price.
  • Evidence:
    • The 424B3 of 10-07 covers:
      • 5.3M and 8.06M resale shares;
      • 11.29M shares from the 6% 2031 notes;
      • up to 26.0M shares from the 4.75% 2032 notes [6].
    • That totals ~50.6M registered shares against 38.3M outstanding (arithmetic). The registrations went effective on 08-14 and 08-21 [6].
    • Class A shares rose from 11.83M (2025-12-31) to 38.34M (2026-09-17), +224% [3,15].
    • The June placement at $68.73 is ~48% under water. The 6% notes convert at $48.24, 35% above today's price [3].
    • Short interest is 2.8M shares (~12% of float, 2.1 days to cover) on 09-30, up from 2.4M on 09-15 [30].
  • Why it matters: holders who are under water can sell into any rally. Convert holders can hedge by shorting, which caps rebounds.
  • 🛡️ Counter-argument: the convert shares only arise far above today's price ($48.24 and $99.66), so most of the 50.6M is theoretical. The largest holder has not sold [15].
  • Severity: Medium

🐻 Bear #9The neocloud funding window is closing

  • The point: The share price fell ~44% from 09-22 to 10-12, mostly on sector news. SharonAI needs capital just as appetite fades.
  • Evidence:
    • Prices: $63.20 on 09-22, $45.29 on 10-07, $37.93 on 10-09, $35.69 on 10-12 [29].
    • 10-08: an FT report put OpenAI's annualised revenue at ~$50B, against ~$68–70B reported earlier. CoreWeave fell ~8% and IREN ~7.7% [32].
    • 10-08/09: Firmus pulled its ~US$5B IPO on weak demand [31].
    • This answers report #1's open question about the cause of the drop: it was sector-wide plus SharonAI's own debt news, not a company-specific event.
  • Why it matters: the ~$3.9B gap (Bear #1) gets more expensive with every week of weak sentiment.
  • 🛡️ Counter-argument: debt markets are still open for contracted GPU capacity: Lambda's $1B at 6.78% on 10-01 and CoreWeave's $3.7B convert on 09-17 [34,36]. Sector scares have reversed quickly before.
  • Severity: Medium High.

🐻 Bear #10Unproven at scale: ~4 MW live, no security certifications

  • The point: It must run tens of megawatts reliably, and NVIDIA's termination clause punishes downtime. It has not yet shown it can.
  • Evidence:
    • Q2 revenue was $1.9M [11].
    • SemiAnalysis (ClusterMAX 3.0, 09-23) placed it in the lowest tier and estimated ~4 MW live (report #1's reading; not re-verified this run) [39].
    • No SOC 2, ISO 27001 or IRAP (Australian government security assessment) certification is claimed in the S-1/A or the 10-Q [5,3].
  • Why it matters: uptime and security decide both NVIDIA termination risk and whether it can sell the 92 MW to enterprises and government.
  • 🛡️ Counter-argument: the first B300 cluster passed customer acceptance [21]. VAST's DataEnclave (09-23) targets regulated customers [9]. Certifications can be bought with time.
  • Severity: Medium

🐻 Bear #11Founder control and a CEO with unresolved Mawson disputes

  • The point: Three founders control ~38% of votes, and the CEO's previous company is still litigating with entities linked to him.
  • Evidence:
    • 136,341 Class B shares carry 160 votes each, 21.8M votes or ~38% [5].
    • Mawson's 10-Q (08-14): Luna Squares Property v. Vertua Property (filed 2024-10-17, Pennsylvania) "remains ongoing". Manning is a director of Vertua Ltd. Mawson also says W Capital is related to Manning [38].
    • These are Mawson's allegations, with no findings.
    • Related-party services continue: First Equity Tax $44K, Bare Media $10K and Shift Advisory $134K in Q2 [3].
  • Why it matters: founder control makes a takeover premium less likely, and an adverse finding would hurt credibility with lenders.
  • 🛡️ Counter-argument: SharonAI says it is party to no material proceedings [3]. The related-party amounts are tiny, and the CEO's consulting deal was terminated without payment.
  • Severity: Low Medium.

🐻 Bear #12Blackwell rental prices are near their peak

  • The point: Five- and six-year contracts are being priced at the top of the cycle for the chips they run on.
  • Evidence:
    • The H100 contract price fell from $6.62 (late 2023) to $2.83 (Feb 2026) [33]. The H100 index was $2.63 on 09-07 [33].
    • The B200 12-month forward price is ~12% below spot [33].
    • NVIDIA's Vera Rubin generation began shipping in its fiscal Q3 (from August 2026; secondary).
  • Why it matters: the take-or-pay periods hold up revenue, but resale above the NVIDIA floor and re-letting after contracts end will happen at lower prices.
  • 🛡️ Counter-argument: H100s still rent at ~$2.6 per hour three years after launch, and inference demand keeps older chips busy.
  • Severity: Low Medium within 36 months, higher after.

Tailwinds NOT in the forecasts

Base for sizing: there is no reliable EPS. FY2026 consensus EPS is −$16.38 and FY2027 is −$7.94, both distorted by convertible fair-value losses, and FY2028 consensus was not found [29]. Impacts are therefore in EBITDA per share. Assumptions (estimates): 38.5M shares and a 60% incremental EBITDA margin. The margin is an estimate, not a company figure.

  • Part of the unsold 92 MW signed at B300-type prices — 30% chance by end-2027
    • Impact on revenue: about +$226M in FY2028, against an FY2028 consensus that was not found. Against FY2027 consensus of $1.54B it would be +15%.
    • Impact on EPS: about +$3.52 EBITDA per share in FY2028.
    • How we got the number (estimate): 20 MW net new (half of an assumed 40 MW, since consensus probably includes some) × ~$11.3M per MW per year (NVIDIA deal rate) × 60% ÷ 38.5M shares ≈ +$3.52.
    • Expected value: 0.30 × $3.52 ≈ +$1.06 per share.
    • What it is: signing new five-year contracts on the 92 MW secured but unsold.
    • Why it's not priced in: the stock trades at 0.89× 2027 consensus, below peers, which implies doubt about even the signed book. Manning would not promise "the same momentum".
    • Trigger to watch: an 8-K with a new contract for 20 MW or more, or a TCV update above $8.8B.
    • Why this probability: demand for B300s is strong, but each new deal also needs ~$39M per MW of capex at a time when the gap is already unfunded.
  • NVIDIA capacity resold above the floor — 22% chance of a material resale by end-2028
    • Impact on revenue: about +$73M in FY2028 (FY2028 consensus not found).
    • Impact on EPS: about +$1.90 EBITDA per share in FY2028.
    • How we got the number (estimate): 10,000 GPUs × 8,760 h × ($4.00 − $2.33) × 50% share kept (the split is undisclosed) ≈ $73M, near 100% incremental margin ÷ 38.5M ≈ +$1.90.
    • Expected value: 0.22 × $1.90 ≈ +$0.42.
    • What it is: third parties buy the 72 MW at market prices, and SharonAI keeps the floor plus its share of the spread.
    • Why it's not priced in: management tells investors not to rely on it [3], and no analyst model discloses it.
    • Trigger to watch: named third-party users on the GB300 clusters, and "shareable revenue" disclosed in the 10-Q.
    • Why this probability: the spread is large today, but the capacity arrives in 2027, forward prices are lower, and the floor falls each year.
  • NVIDIA build financed at ≤7% — 25% chance by mid-2027
    • Impact on revenue: none.
    • Impact on EPS: about +$1.62 per share of pre-tax earnings in FY2028 from lower interest. It is pre-tax, and SharonAI pays no tax while loss-making.
    • How we got the number (estimate): $2.5B of debt × 2.5 points saved versus ~9.5% = $62.5M ÷ 38.5M ≈ +$1.62.
    • Expected value: 0.25 × $1.62 ≈ +$0.41.
    • What it is: a multi-billion facility priced near Lambda's 6.78%, secured on NVIDIA's floor.
    • Why it's not priced in: the only data point is the 9.95% first facility.
    • Trigger to watch: an 8-K with named bank lenders and a fixed rate of 7% or less.
    • Why this probability: NVIDIA's credit helps, but the program pause, the Firmus failure and a weak operating record argue for high-yield pricing.
  • Takeover by a larger operator or infrastructure fund — 8% chance within 36 months
    • Impact on revenue and EPS: not quantifiable. A takeover changes the price, not the earnings.
    • Expected value: not quantifiable.
    • What it is: a buyer wanting Australian AI capacity and an NVIDIA contract buys the company below its cash cost of build.
    • Why it's not priced in: analysts model it stand-alone.
    • Trigger to watch: a 13D from a strategic buyer, or a change-of-control review event under the facility.
    • Why this probability: dual-class founder control and ~$6B of commitments make a deal hard.

Headwinds NOT in the forecasts

  • Equity raise of ~$750M near or below today's price — 60% chance by end-2027
    • Impact on revenue: none directly; it funds the revenue already in consensus.
    • Impact on EPS: about −$9.5 FY2027 EBITDA per share (−39%).
    • How we got the number (estimate): consensus FY2027 revenue $1.54B × 60% = $924M EBITDA. That is $24.0 per share on 38.5M shares, or $14.5 per share on 63.5M shares (+25M new shares at $30), a difference of −$9.5.
    • Expected value: 0.60 × −$9.5 ≈ −$5.7.
    • What it is: equity, a low-priced convertible or a preferred issue to fill part of the ~$3.9B gap.
    • Why it's not priced in: analyst targets of $80–124 imply no dilution below $36 [29].
    • Trigger to watch: an S-3 or ATM program, a waiver of the 6% notes' equity lock, or a convertible announcement.
    • Why this probability: project debt rarely covers 100% of a ~$4B build. The first loan's ~49% loan-to-value suggests lenders will fund only about half.
  • Build delays shorten the NVIDIA contract — 45% chance that the average phase slips 3 months or more
    • Impact on revenue: about −$204M in FY2027 (−13% vs consensus of $1.54B).
    • Impact on EPS: about −$4.2 FY2027 EBITDA per share.
    • How we got the number (estimate): $817M ÷ 4 = $204M of floor revenue lost permanently × 80% decremental margin (costs keep running) ÷ 38.5M ≈ −$4.2.
    • Expected value: 0.45 × −$4.2 ≈ −$1.9.
    • What it is: NEXTDC hand-over, GPU delivery or acceptance tests slip, and because all phases end on the same date, the lost months are never recovered.
    • Why it's not priced in: consensus models the contract as full-term.
    • Trigger to watch: NEXTDC phase dates in the Q3 10-Q, a GB300 delivery update, a TCV reduction.
    • Why this probability: five phases in five months for a company that slipped its first cluster by about two months.
  • NVIDIA changes or ends its floor — 12% chance within 36 months
    • Impact on revenue: about −$400M in FY2028 (consensus not found; −26% against the FY2027 consensus of $1.54B as a scale reference).
    • Impact on EPS: about −$10.4 FY2028 EBITDA per share.
    • How we got the number (estimate): the $817M floor is replaced by market resale for about half the capacity, so −$400M × ~100% decremental margin ÷ 38.5M ≈ −$10.4.
    • Expected value: 0.12 × −$10.4 ≈ −$1.25.
    • What it is: a termination over uptime breaches, or a renegotiation after NVIDIA's program review.
    • Why it's not priced in: analysts treat the $4.9B as firm.
    • Trigger to watch: NVIDIA 10-Q wording on AI-cloud commitments, a SharonAI 8-K amending the agreement, an outage disclosure.
    • Why this probability: the contract is signed and NVIDIA confirms the program, but uptime termination rights are real for a first-time operator.
  • ESDS fails to pay or renegotiates — 20% chance within 36 months
    • Impact on revenue: about −$190M in FY2027 (−12% vs $1.54B).
    • Impact on EPS: about −$3.0 FY2027 EBITDA per share.
    • How we got the number (estimate): $252M a year lost, half re-let after 6 months, so ~−$190M × 60% ÷ 38.5M ≈ −$3.0.
    • Expected value: 0.20 × −$3.0 ≈ −$0.60.
    • What it is: ESDS cannot pass through ~$250M a year to its own clients, or sanctions issues block payment.
    • Why it's not priced in: TCV includes the full $1.26B.
    • Trigger to watch: an ESDS billing start in the Q3 10-Q, a facility review-event notice, and ESDS's own quarterly results on Indian exchanges.
    • Why this probability: ESDS's fresh IPO cash and its rising share price help, but the commitment is ~4.7× its revenue.
  • FY2026 revenue far below consensus — 50% chance
    • Impact on revenue: about −$58M in FY2026 (−49% vs $118.6M).
    • Impact on EPS: about −$0.9 FY2026 EBITDA per share.
    • How we got the number (estimate): ESDS and APAC run at about half rate in Q4, so ~$60M against $118.6M; −$58M × 60% ÷ 38.5M ≈ −$0.9.
    • Expected value: 0.50 × −$0.9 ≈ −$0.45.
    • What it is: a Q4 ramp slower than consensus assumes.
    • Why it's not priced in: consensus is still at $118.6M [29].
    • Trigger to watch: Q3 results and Q4 commentary.
    • Why this probability: the date was not confirmed on the Q2 call, and consensus needs both contracts live on time.

Investment Risk Score: 73/100 (high)

# Factor Weight Measured value Score Why
1 Liquidity & funding runway 15% Cash $1.86B (06-30) + $356M facility vs ~$6.1B commitments → ~$3.9B unfunded; Q2 underlying burn ~−$113M operating + −$268M advances 72 Operating cash lasts years, but the commitments due by mid-2027 don't fit. No going-concern language.
2 Leverage 10% Debt ~$1.41B ($1.05B converts + $356M facility) vs cash $1.86B; Q2 adj. EBITDA $0.6M; facility 9.95% with parent guarantee 62 Net cash on paper, but the cash is committed and EBITDA is ~zero, so the debt can't be serviced from operations yet.
3 Dilution 10% Class A 11.83M → 38.34M (+224%, Dec-25 → Sep-26); converts 14.3M (up to 24.1M max) + 1.1M options + plan evergreen 95 Far above the +25% anchor, with a large overhang.
4 Profitability 10% Q2 revenue $1.9M, loss from operations −$24.8M (−1,283% margin); H1 FCF ≈ −$430M excluding deposits 85 Pre-scale, deep losses.
5 Revenue quality & concentration 10% NVIDIA ~55% of TCV, ESDS ~14%; 2025 top 3 = 82% of revenue; no RPO disclosed 78 Above 35% with one counterparty, but multi-year take-or-pay.
6 Execution & track record 10% 1H-26 cluster slipped to Aug; $470M run-rate dropped; open material weakness; 10-Q/A; PR vs 10-Q EPS mismatch; three auditors 82 Material weakness plus restated comparisons and missed timing.
7 Competitive position 10% ~4 MW live; ClusterMAX lowest tier; no SOC 2/ISO; 212 MW secured; one of three Australian NVIDIA cloud partners 65 Contracted demand, but an easy-to-copy product with no proven operations.
8 Cyclicality & macro 10% Rental prices: H100 −57% in ~2.5 years; B200 forward −12%; stock −44% in 3 weeks on sector news 68 Take-or-pay softens it; resale, re-letting and funding are all cyclical.
9 Regulatory, legal & geopolitical 5% ESDS top client a sanctioned Russian bank (22%, FY25 H1); US export rules; AEMC large-load rule late Oct; Mawson litigation 55 Material but contained; no proceedings against SharonAI itself.
10 Valuation & expectations 10% Market cap $1.37B = 0.89× FY27 consensus; 6/6 Buy, targets $80–124; short ~12% of float 58 Cheap multiple, but consensus assumes perfect timing plus new sales.

Overall: weighted average = 72.85 → 73. Change vs last: +2 against the old 6-area 71, and not a like-for-like comparison. The v3 method measures dilution (95) and profitability (85) directly, which lifts the score. Leverage (62) and valuation (58) pull it down because the June cash is still on the balance sheet.

Ranking check: SHAZ now ranks as the riskiest holding in the ledger, just above GRRR (72), because more of its value is still unbuilt and unfunded. GRRR's dilution is mechanically scheduled, while SHAZ must raise ~$3.9B. It sits clearly above IREN (69), which has live operations and a Microsoft contract already billing.

Conclusion

  • What you are really betting on: that a team running ~4 MW today can hand over ~200 MW of NVIDIA clusters on time from Q4 2026 to August 2027, and fund ~$3.9B of it without heavy dilution below today's price.
  • The 2–3 things that will decide the outcome:
    1. The financing for the ~$4B NVIDIA build: rate, loan-to-value and how much equity it needs.
    2. Delivery dates. Every month of NVIDIA slippage is ~$68M of floor revenue lost for good.
    3. Whether ESDS and the APAC contract bill in full in Q4.
  • Signposts:
    • Confirming the bull case:
      • Q3/Q4 revenue showing ESDS and APAC billing at a run-rate of ~$100M a quarter;
      • a multi-billion NVIDIA-backed facility at ≤7%;
      • a new contract on the 92 MW at $4+ per GPU-hour;
      • a SOC 2/ISO certification.
    • Breaking the bull case:
      • equity or a low-priced convertible below ~$36;
      • a NEXTDC phase or GB300 delivery slip;
      • any review event, ESDS billing delay or TCV cut;
      • an NVIDIA amendment or an outage disclosure.
  • New since the last report:
    • NVIDIA's contract phases all end on the same date, and NVIDIA can terminate for frequent unavailability. Delay and downtime are now contract risks, not just timing risks.
    • The ESDS prospectus confirms a sanctioned Russian bank was its largest client (22%), and ESDS's yearly commitment is ~4.7× its revenue.
    • The September–October fall was sector-driven (the OpenAI revenue report and Firmus's failed IPO). The $470M run-rate claim is confirmed as said on the Q1 call and then dropped at Q2.

Open questions & data gaps

  • Facility drawdown and Facility A trigger: not found. Drawn amount, the redacted loan-to-value levels and the escrow amount all decide when parent recourse ends. The Q3 10-Q is the next source.
  • Commitment schedule by year: not found. Only the ~$6B total is disclosed. The timing of the ~$4B NVIDIA spend is the key runway variable.
  • NVIDIA contract: the floor schedule, revenue-share split and the uptime threshold are not public, and SharonAI has not said whether the August pause affected it. This matters most.
  • ESDS: whether billing started, whether the $140M of LCs/guarantees exists in the latest filings, and the bank's name (Gazprombank per Bleecker, not named in the prospectus).
  • Share count: 35.27M (proxy record date) vs 35.67M (06-30) vs 38.34M (09-17) are not reconciled. The maximum convert shares are 14.3M, 24.1M or 26.0M depending on the document.
  • Consensus assumptions: MW live, $/MW and the FY2028 numbers are behind paywalls.
  • Live MW and the ClusterMAX tier: not re-verified this run.
  • Q3 earnings date: not announced; estimated early to mid-November.
Sources
  1. SharonAI Holdings Form 10-K FY2025 — SEC EDGAR — 2026-03-31 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226014068/form10-k.htm
  2. SharonAI Form 10-Q, quarter ended 2026-03-31 — SEC EDGAR — 2026-05-15 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226023769/form10-q.htm
  3. SharonAI Form 10-Q, quarter ended 2026-06-30 — SEC EDGAR — 2026-08-06 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226036377/form10-q.htm
  4. SharonAI Form 10-Q/A Amendment No. 1 — SEC EDGAR — 2026-08-11 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226037202/form10qa.htm
  5. SharonAI Form S-1/A Amendment No. 2 — SEC EDGAR — 2026-08-20 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226039475/forms-1a.htm
  6. Prospectus supplements (424B3) — SEC EDGAR — 2026-10-07 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226046083/form424b3.htm
  7. Form 8-K and Ex. 10.1 Syndicated Facility Agreement (US$356M) — SEC EDGAR — 2026-10-01 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226045368/form8-k.htm ; https://www.sec.gov/Archives/edgar/data/2068385/000149315226045368/ex10-1.htm
  8. Sharon AI Enters Into GPU-Backed Debt Facility (Ex. 99.1) — SEC EDGAR — 2026-10-01 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226045368/ex99-1.htm
  9. Form 8-K Item 4.01 (HoganTaylor dismissed, EY engaged) and Ex. 16.1; 8-K VAST DataEnclave — SEC EDGAR — 2026-09-23/24 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226044154/form8-k.htm ; https://www.sec.gov/Archives/edgar/data/2068385/000149315226043802/form8-k.htm
  10. Form 8-K, NVIDIA Master Cloud Services Agreement (up to $4.88B) — SEC EDGAR — 2026-06-12 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226028370/form8-k.htm
  11. Sharon AI Reports Second Quarter 2026 Results (Ex. 99.1) — SEC EDGAR — 2026-08-06 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226036254/ex99-1.htm
  12. Sharon AI Reports First Quarter 2026 Results — Business Wire — 2026-05-15 — https://www.businesswire.com/news/home/20260515327013/en/Sharon-AI-Reports-First-Quarter-2026-Results
  13. SharonAI Q1 2026 earnings call transcript — Investing.com — 2026-05-15 — https://ca.investing.com/news/stock-market-news/earnings-call-transcript-sharonai-q1-2026-sees-strategic-growth-postipo-93CH-4643946
  14. SharonAI Q2 2026 earnings call transcripts (call held 2026-08-06) — Investing.com / Motley Fool — 2026-08-06 / 2026-08-13 — https://www.investing.com/news/transcripts/earnings-call-transcript-sharonai-q2-2026-miss-offsets-strong-pipeline-growth-93CH-4844537 ; https://www.fool.com/earnings/call-transcripts/2026/08/13/sharonai-shaz-q2-2026-earnings-call-transcript/
  15. Schedule 13D and 13D/A, Situational Awareness LP; Form 4 — SEC EDGAR — 2026-08-28 / 2026-09-17 — https://www.sec.gov/Archives/edgar/data/2068385/000093583626000500/primary_doc.xml ; https://www.sec.gov/Archives/edgar/data/2068385/000093583626000494/form4.xml
  16. Schedule 13G/A, Oaktree Capital Management — SEC EDGAR — 2026-08-10 — https://www.sec.gov/Archives/edgar/data/2068385/000114036126032029/primary_doc.xml
  17. Schedule 13G/A, Millennium — SEC EDGAR — 2026-08-14 — https://www.sec.gov/Archives/edgar/data/2068385/000110465926097339/primary_doc.xml
  18. Form 8-K and Ex. 99.1, $350M 6% convertible notes (founder lock-up to 2027-03-31; equity restriction) — SEC EDGAR — 2026-04-28 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226019260/ex99-1.htm
  19. Form 8-K, First Supplemental Indenture (2031 notes covenants removed) — SEC EDGAR — 2026-08-26 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226040152/form8-k.htm
  20. Form 8-K, ~$900M equity and $700M 4.75% convertible notes — SEC EDGAR — 2026-06-17 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226029005/form8-k.htm
  21. Sharon AI Successfully Delivers AI Cloud Deployment (Ex. 99.1) — SEC EDGAR — 2026-08-20 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226039364/ex99-1.htm
  22. Form 8-K, COO appointment (Burns) and Leece Deed of Release — SEC EDGAR — 2026-08-27 / 2026-09-11 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226040426/form8-k.htm ; https://www.sec.gov/Archives/edgar/data/2068385/000149315226042434/form8-k.htm
  23. Form 8-K, CFO appointment (Goel) — SEC EDGAR — 2026-07-24 — https://www.sec.gov/Archives/edgar/data/2068385/000149315226034569/form8-k.htm
  24. NVIDIA Form 10-Q, quarter ended 2026-07-26 — SEC EDGAR — 2026-08 — https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm
  25. SharonAI stock falls 8% after Nvidia pauses revenue-sharing deals — Investing.com — 2026-08-28 — https://ca.investing.com/news/stock-market-news/sharonai-stock-falls-8-after-nvidia-pauses-revenuesharing-deals-93CH-4821341 ; Nvidia pauses AI cloud revenue-sharing deals — The Next Web — 2026-08-28 — https://thenextweb.com/news/nvidia-pauses-ai-cloud-revenue-sharing-deals
  26. ESDS Software Solution Draft Red Herring Prospectus — ESDS / BSE — 2025-03-30 — https://www.esds.co.in/policies_document/DRHP/ESDS_DRHP.pdf
  27. ESDS IPO details and listing — Kotak Neo / IIFL / Bajaj Broking — 2026-08 / 2026-09-08 — https://www.kotakneo.com/news/ipos/esds-software-ipo-august-28-price-band/ ; https://www.indiainfoline.com/news/companies/esds-share-price-hits-20-upper-circuit-again-shares-up-200-from-ipo-issue-price ; https://www.bajajbroking.in/blog/esds-software-solution-limited-ipo
  28. SharonAI (SHAZ) short report — Bleecker Street Research (discloses short position) — 2026-04-30 — https://www.bleeckerstreetresearch.com/research/shaz
  29. SHAZ forecast, statistics and price history; CRWV/NBIS/IREN statistics — stockanalysis.com (S&P data) — 2026-10-12 — https://stockanalysis.com/stocks/shaz/forecast/ ; https://stockanalysis.com/stocks/shaz/statistics/
  30. SHAZ short interest history (FINRA data) — curvedtrading — 2026-10-12 — https://curvedtrading.com/stocks/shaz/short-interest/
  31. Nvidia-backed Firmus scraps US$5bil Australian IPO on poor demand — Reuters via New Straits Times — 2026-10-09 — https://www.nst.com.my/amp/business/corporate/2026/10/1551088/nvidia-backed-firmus-scraps-us5bil-australian-ipo-poor-demand
  32. Nvidia, Oracle, CoreWeave and other AI stocks sink on OpenAI revenue report — CNBC — 2026-10-08 — https://www.cnbc.com/2026/10/08/open-ai-revenue-nvidia-oracle-coreweave.html
  33. B200 rental price August 2026 update — Silicon Data — 2026-09-24 — https://www.silicondata.com/blog/b200-rental-price-august-2026-update ; Data center GPU pricing 2026 — IntuitionLabs — 2026-07 — https://intuitionlabs.ai/articles/data-center-gpu-pricing-2026
  34. Lambda closes $1 billion senior secured fixed rate financing — Lambda — 2026-10-01 — https://lambda.ai/blog/lambda-closes-1-billion-senior-secured-fixed-rate-financing
  35. Nebius raises $775 million in first secured debt financing — Nebius — 2026-07-17 — https://nebius.com/newsroom/nebius-raises-775-million-in-first-secured-debt-financing-to-accelerate-global-buildout
  36. CoreWeave senior notes pricing; CoreWeave $3.7B convertible notes — Business Wire / CoreWeave IR — 2026-06-11 / 2026-09-18 — https://secure.businesswire.com/news/home/20260611950890/en/CoreWeave-Announces-Pricing-of-%241.25-Billion-of-Senior-Notes-and-%E2%82%AC2-Billion-of-Senior-Notes ; https://investors.coreweave.com/news/news-details/2026/CoreWeave-Prices-Upsized-3-7-Billion-Convertible-Senior-Notes-Offering/
  37. NEXTDC FY26 results release and presentation — ASX — 2026-08-27 — https://announcements.asx.com.au/asxpdf/20260827/pdf/073c04gt4ts40y.pdf ; https://announcements.asx.com.au/asxpdf/20260827/pdf/073c08w3kby6d5.pdf
  38. Mawson Infrastructure ("Big Digital") Form 10-Q, quarter ended 2026-06-30 — SEC EDGAR — 2026-08-14 — https://www.sec.gov/Archives/edgar/data/1218683/000121390026089986/ea0299976-10q_bigdigital.htm
  39. ClusterMAX 3.0 — SemiAnalysis — 2026-09-23 — https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard
  40. Compass Point reiterates SharonAI on Nvidia deal — Investing.com — 2026-06 — https://www.investing.com/news/analyst-ratings/compass-point-reiterates-sharonai-stock-rating-on-nvidia-deal-93CH-4741894 ; Northland initiates at Outperform, $95 — MarketBeat — 2026-09-18 — https://www.marketbeat.com/instant-alerts/analyst-northland-securities-begins-coverage-on-sharonai-holdings-inc-class-a-common-stock-nasdaq-shaz-2026-09-18/
  41. OFAC SDN list, Gazprombank entry — U.S. Treasury — accessed 2026-10-12 — https://sanctionssearch.ofac.treas.gov/Details.aspx?id=17016
  42. Firmus could slash share price by nearly 20% or pull IPO — SmartCompany — 2026-10-07 — https://www.smartcompany.com.au/artificial-intelligence/firmus-could-slash-share-price-by-nearly-20-or-pull-ipo/ ; Firmus pulls ASX IPO after 48 hours — Crypto Briefing — 2026-10 — https://cryptobriefing.com/firmus-pulls-asx-ipo-after-48-hours/
  43. Why is SharonAI Holdings stock sliding today — Investing.com — 2026-09-23 — https://www.investing.com/news/stock-market-news/why-is-sharonai-holdings-stock-sliding-today-93CH-4913390
  44. Australian electricity and gas markets July 2026 update (AEMC large-load standard timing) — Ashurst — 2026-07 — https://www.ashurstperkinscoie.com/en/insights/australian-electricity-and-gas-markets-july-2026-update/
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