2026-10-02 — Nvidia touches its first record since May, then closes just short of one

• Nvidia shares touched their first record high since May on Friday, up as much as 3%, then closed up 1.3%, just short of the May record, as its market value neared $6 trillion, Bloomberg reported Bloomberg’s first dispatch was at about 9:47am ET and its headline says the value “nears $6 trillion”; Quartz puts the value at around $5.7 trillion and attributes the run to the $150 billion buyback increase Nvidia’s board approved on Monday, September 28, and the company’s agent-security system. Bloomberg’s closing account says the stock trimmed its gains after a two-month selloff that wiped out more than $1 trillion of market value, leaving its value at roughly $5.6 trillion. (Bloomberg, Bloomberg close, via Economic Times, Quartz)

2026-09-23 — Supermicro says it is shipping Vera Rubin NVL72 racks

• Supermicro announced on 09-23 that it is now shipping Nvidia Vera Rubin NVL72 racks, integrated with its liquid-cooling stack, with customers able to order “Scalable Units” of production-ready systems. The press release gives no volumes or customer names, so it is a vendor claim about availability rather than an order-book number, but Rubin availability is what the ~$1T Blackwell-plus-Rubin book depends on. (PR Newswire)

2026-09-17 — Huang escalates his own growth claim from 70% revenue to a doubling of chip sales volume, made at a UK AI-safety summit alongside the frontier labs

• Jensen Huang told media ahead of a summit with King Charles III — also attended by representatives from Google DeepMind, OpenAI and Anthropic to discuss AI safety — that Nvidia will sell twice as many chips next year as this year, citing AI investment demand “in almost every single country that we’re in.” This is a volume claim, one notch up from the ~70% revenue growth claim this thread logged from him on 09-11 at Goldman’s Communacopia conference (a ~$680bn implied revenue figure on a ~$400bn base); Huang made both claims within a week of each other. Separately on 09-14, Huang told Trump directly “we’re not going to let [an AI slowdown] happen,” a demand-side reassurance made the same week Anthropic’s Dario Amodei publicly called for the industry to slow model-capability development (see AI Bear Turn) — Huang is the clearest industry voice pushing back on that call. Both claims rest on conference/media remarks, not an Nvidia IR filing or transcript. (CNBC, TechCrunch)

2026-09-11 — Jensen Huang tells Goldman’s Communacopia conference he’s confident Nvidia can grow revenue 70% next fiscal year

• ⚠️ Jensen Huang told the Goldman Sachs Communacopia + Technology conference he is “confident” Nvidia can grow revenue 70% year-over-year next fiscal year — roughly $680bn on a current-year base near $400bn — grounding the claim in AI-compute demand still exceeding supply and saying Nvidia is turning away business it cannot fill. He separately flagged cybersecurity as the next major AI growth vertical after training and inference. This is a fresh, on-the-record demand-side number from Nvidia’s own CEO to set against this thread’s existing ~$1 trillion Blackwell+Rubin backlog aggregate, but it rests on conference coverage cross-read across three outlets, not an Nvidia IR document or transcript. (TechCrunch)

2026-09-07 — NYT: blacklisted Inspur kept buying Nvidia chips, including Blackwell, through a part-owned subsidiary

• ⚠️ The New York Times reported that Inspur, blacklisted by the US in 2023, kept buying advanced Nvidia chips — including Blackwell parts — through Aivres Systems, a nominally independent Silicon Valley subsidiary it part-owns (roughly one-third). The investigation puts $5.6bn of hardware moving through that channel April 2024 → February 2026 — a live example of this thread’s demand-side question (who is actually buying, through what channel) running into export-control evasion. The NYT text itself is paywalled and was not read directly; the $5.6bn figure comes from secondary outlets summarising it, so treat it as reported-not-verified. A separate $2bn Megaspeed/Malaysia claim in the same cluster appears to recycle an October 2025 story and is excluded here pending a dedupe check. ⚠️ No direct NYT link located; recorded from secondary summaries.

2026-09-03 — The deal closes the arc this thread tracked since 08-26 — and the confirmed total is lower than the rumor this thread carried on 09-01

• Nvidia’s own 8-K puts the total deal value at $12.93bn ($11.9bn to stockholders + up to $1.0bn employee retention) — resolving this thread’s 08-26 “agreed” report and its 09-01 update, and correcting the ~$14bn total the 09-01 entry carried. The 09-01 entry (sourced to Bloomberg) read the reported $12.9B figure as the base price with a rumored $1B retention package layered on top, putting the total at “~$14B.” The primary filing shows the two instruments actually sum to $12.93bn, not ~$14bn — the earlier escalation did not hold up against Nvidia’s own disclosed structure. See the correction below. (SEC EDGAR 8-K, TechCrunch)

2026-09-01 — Bloomberg reports the Hugging Face deal could close this week at a higher total value

• Bloomberg reported Tuesday evening that Nvidia is nearing a roughly $14B acquisition of Hugging Face, with the agreement possibly reached “as soon as this week.” This updates, rather than repeats, the 08-27 entry already on this thread (“agreed… still unconfirmed” at ~$12.9B): the reported base price is unchanged at $12.9B, but the total is now put at ~$14B once a reported $1B employee-retention package is included, and the timeline has moved from an open-ended “agreed” to a specific “this week.” Both companies again declined to comment; no final agreement is confirmed by either side, and terms/timing could still change. This is a demand-side/ecosystem move (what Nvidia is buying, not what it’s selling), but the thread has carried it since 08-26 for the same reason it’s flagged again here: the store of open-source models Nvidia would own bears on how “the order book” gets filled downstream. (Bloomberg, Investing.com syndication, timestamped 2026-09-01 8:56 PM, Seeking Alpha)

2026-08-27 — The guide is $108.0bn, and that is the number the tape traded

• Nvidia guided Q3 FY2027 revenue to $108.0bn, plus or minus 2%, per its own release, on top of $96.2bn of actual Q2 FY2027 revenue, up 106% year over year. The stock closed +8.74% at $227.98. ⚠️ The widely-quoted $104.2bn “consensus” is a media-reported analyst aggregate, not an Nvidia figure, and the two should not be set against each other as if they came from the same kind of source. This entry also settles the prior day’s hedge: the map could only say “roughly 7-8%” live, and the close is now on the record. (Nvidia investor relations)

2026-08-27 — Reporting firms to an agreed Hugging Face deal, and the rally carries

• CNBC, citing a person familiar, reported that Nvidia has AGREED to acquire Hugging Face for about $12.9B — an escalation from the previous evening’s “in talks.” Reporting notes an earlier $500M offer was rejected and that Hugging Face was valued at $4.5B in 2023. It would be Nvidia’s largest acquisition ever, against $7B for Mellanox. ⚠️ Still unconfirmed by either company, sourced entirely to people familiar, and the original reporting said it could fall through. (CNBC, The Information)

• Nvidia rose roughly 7-8% in Thursday’s regular session, pulling Marvell (+5.8%) and Micron (+4.5%) with it. The precise intraday figure varies by source and sampling minute and is not asserted. (CNBC)

2026-08-26 — The print lands, and the filing beside it is the bigger document

• Nvidia reported Q2 FY2027 revenue of $96.2B (+106% YoY) with data-centre revenue of $89.0B (+117%), beating the $91-95bn consensus, and guided Q3 to $108.0B ±2% — while assuming zero China data-centre compute revenue. GAAP EPS $2.46, non-GAAP $2.22, gross margin 75.0%; Q3 margin guided to 74.0% ±50bps on rising HBM and DRAM costs. The demand side of this thread has run for months on Huang’s own ~$1T aggregate; this is the first quarter where the guide alone clears $100B. ⚠️ Naming correction carried across the record: this was Nvidia’s fiscal 2027 Q2, not FY2026 — a label wrong in every digest since 08-20. (SEC 8-K, NVIDIA IR)

• The 10-Q filed the same day disclosed guarantees capped at $105B tied to an affiliate of OpenAI Group PBC, $3.5B more in lease guarantees for AI cloud partners, and a $25B committed equity pool for AI labs and infrastructure financiers. The call added $500bn+ of planned third-party capital through financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR. This thread’s circular-financing angle stops being an inference here — it is now filed, dated and capped in a document with liability attached. (SEC EDGAR, NVDA 10-Q)

• Receivables concentration widened to five customers at 72% of the balance (22/14/13/11/10%), from three at 56% (25/18/13%) two quarters earlier. More names carrying large balances, only modestly less concentration overall — and one of those counterparties is the affiliate the $105B of guarantees is written to. (SEC EDGAR, NVDA 10-Q)

• Bloomberg reported at 20:56 ET that Nvidia is in talks to buy Hugging Face for roughly $13B. Unconfirmed by either company. Noted here because this thread tracks what Nvidia sells rather than what it buys — the acquisition belongs to a thread that does not exist yet, offered as a candidate 08-27. (Bloomberg)

• Nvidia CFO Colette Kress told analysts on the 08-26 earnings call that demand from the AI labs Nvidia supports with its own balance sheet will contribute toward roughly a quarter of its business next year, and that Nvidia has put nearly $50 billion into the frontier labs that buy its chips. Kress acknowledged that some would call the arrangement circular financing, and said Nvidia sees it differently. She also named the financing partners (Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR) for platforms meant to raise more than $500 billion of outside capital, and said the first phase of a site secured with SB Energy supports 4.25 gigawatts for OpenAI, whose existing and planned commitments she put at about 12 gigawatts of Nvidia compute through 2030. This turns this thread’s circular-financing question into a stated share of demand: a management-quantified fraction of the order book depends on customers Nvidia is funding, on top of the $105bn of guarantees and the receivables concentration already logged from the 10-Q the same day. (AI News)

2026-08-25 — Wall Street puts a first dollar figure on the Rubin ramp, a day before the print that will test it

• Morgan Stanley estimates Nvidia’s new Vera Rubin chips could contribute nearly $9B in sales in the fiscal third quarter ending October, the first analyst dollar figure attached to the Blackwell-to-Rubin transition this thread has recorded. Wall Street consensus separately expects Nvidia to guide third-quarter revenue up 82.8% year-on-year to $104.2B when it reports Wednesday. The order book’s demand side has run for months on Huang’s own ~$1T aggregate figure; this is the first outside, per-product estimate of how much of it Rubin specifically is expected to deliver in the near term. (Reuters, via Yahoo Finance)

• OpenAI and Broadcom published Jalapeño’s first performance benchmarks at Hot Chips 2026, claiming better performance-per-watt than Nvidia’s flagship. The inference ASIC the two announced in June was disclosed at 700W TDP, with a 128-chip pod at 1.7 exaflops in 4-bit and 27.5TB of HBM4, and claimed advantages up to 1.9x throughput-per-kilowatt and 3.6x lower latency on SemiAnalysis’s InferenceX benchmark; small-scale deployment inside OpenAI is targeted for end-2026, broader rollout 2027. The June item was a partnership announcement — this is the first evidence the silicon works. (OpenAI, TechCrunch)

2026-08-24 — SpaceXAI adopts Nvidia’s Vera CPU and the NVL72 rack, including for orbit

• Nvidia announced that SpaceXAI is adopting the Vera Rubin NVL72 rack architecture — Vera CPUs plus Rubin GPUs — for gigawatt-scale terrestrial data centers and for the planned first-generation Starmind AI1 satellite. This extends the exclusive Starmind arrangement already on this timeline from 2026-08-05: that entry established the deal, this one names the silicon. ⚠️ Nvidia’s own release labels the orbital piece planned and carries a forward-looking-statements disclaimer; reported prototype testing is early 2027, so nothing has flown. The “first-ever standalone CPU” framing circulating in secondary coverage is stale — the standalone Vera CPU launched separately in June 2026. (NVIDIA Newsroom, The Register)

2026-08-10 — Nvidia recruits Wall Street to finance $500B of demand-side buildout

• Nvidia partnered with six asset managers — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR — on financing platforms designed to raise over $500B in third-party capital for AI infrastructure, with Jensen Huang saying Nvidia itself can backstop up to $125B (25%) of the resulting deals — MOUs, not final agreements yet. Not a chip order, but the demand-side counterpart to this thread’s whole story: outside capital now being recruited at Wall Street scale specifically to fund the infrastructure that generates Nvidia’s order book, treating compute as a lend-against asset class rather than one-off vendor financing. Nvidia shares fell on the announcement. Full detail on Circular Financing (sev=major there). (NVIDIA Newsroom, CNBC)

2026-08-07 — Nvidia takes a $3B stake in the power layer under Stargate

• Nvidia will invest up to $3B in Lancium, the Blackstone-backed power developer behind Stargate’s Abilene, Texas campus — $2B initial for a ~20% stake, up to $1B more on grid milestones, valuing Lancium at roughly $10B. Not a chip order, but the same customer-financing logic one layer down: Nvidia now holds equity in the power infrastructure its own GPU customers depend on. Cross-refs Circular Financing, Power Buildout. (The Information)

2026-08-05 — A new demand category: SpaceX commits exclusively to Nvidia for AI compute in orbit

• SpaceX unveiled Starmind AI1, an exclusive deal to run Nvidia Rubin GPUs and Vera CPUs as orbital AI compute — each satellite carries a full Rubin NVL72 rack (72 GPUs), stands ~20m tall with a 75m solar wingspan, draws ~210kW average/~250kW peak power, targets a 2027 launch, and SpaceX has floated a constellation scaling toward 1M satellites as a “distributed AI supercomputer in space.” AMD shares fell on the exclusivity news. (Interesting Engineering; multiple outlets 08-04/08-05)
• 💡 A genuinely new category on this thread’s demand side — every prior entry here is terrestrial hyperscaler/lab demand; this is the first orbital compute customer, at satellite counts that dwarf any single datacenter campus if the constellation plan holds. Cost and firm order size are not yet disclosed — treat the “1M satellites” figure as a stated ambition, not a backlog number.

2026-07-24 — Opened + crawled

• Thread opened.
• Verdict: the strongest demand book ever reported — ~$1T in Blackwell+ Rubin orders through 2027 — but it’s Huang’s own aggregate figure, not an audited backlog, and the flashiest pillar (the $100B OpenAI deal) COLLAPSED in early 2026. The binding constraint has shifted from demand to supply (an HBM4 drought). ⚠ WebSearch budget was exhausted (9-crawl fan-out) — built from targeted fetches; a few figures are soft.

← Backstory (crawl 2026-07-24 → artifacts/findings/nvidia-order-book-2026-07-24.md)

2026-05-13 — First-ever $5.5T

• Nvidia hit $5.5T market cap — the first company ever — on news Huang would join Trump’s China trip (~$5T as of late July). (Forbes)

2026-04-09 — The constraint flips to supply

• A Rubin supply squeeze — HBM4 shortage pushed SK Hynix’s ramp to Q3'26; TrendForce cut its 2026 Rubin shipment share 29%→22%. Demand isn’t the bottleneck anymore; memory is. ⚠ reported (TrendForce). (SDxCentral)

2026-03-16 — The $1T claim

• GTC: Huang says ~$1T in Blackwell+Rubin orders through 2027 — up from the $500B-through-2026 figure a year prior. It’s a demand/visibility number, NOT audited RPO/backlog — no public per-customer breakdown exists. (TechCrunch)

2026-03-04 — The $100B OpenAI deal dies

• The OpenAI–Nvidia $100B / 10GW partnership is dead — Huang confirms it’s not proceeding (citing OpenAI’s IPO plans); FT reported the abandonment in Feb. Corrects the circular-financing picture: Nvidia’s actual OpenAI exposure is the ~$30B equity stake from the $110B round, not the $100B LOI. Cross-refs Circular Financing. (Gizmodo)

2025-12 — The backlog + the circular stakes

• ~3.6M-unit Blackwell backlog from cloud providers; B200/GB200 sold out through mid-2026 (Microsoft the lead buyer, Meta >$100B 2026 capex). ⚠ trade/analyst-sourced. (FinancialContent)
• Nvidia’s own 10-Q lists customer investments — OpenAI up to $100B, Anthropic $10B, Intel $5B — Nvidia putting capital into its own buyers, the circular-financing critique made concrete (CFO: no definitive OpenAI agreement). (Fortune)