Tag: IPO

Anthropic IPO Faces Valuation Pushback Ahead of Thanksgiving Window (2026)

Takeaway: Anthropic is trying to go public in a window that, according to the Wall Street Journal, is the week before the U.S. Thanksgiving holiday (Thursday, November 26, 2026). CNBC has reported that the company plans to list on Nasdaq under the ticker ANT, and that as of November 6 it was expected to kick off its roadshow the following week. What is not settled — and what several major investors have been telling reporters — is the price. After earlier talk of a valuation around $2 trillion, more recent discussion centered on about $1.5 trillion and a raise of up to about $100 billion. People who plan to participate have since told the Journal that $1.5 trillion may need to come down, with one saying the number could be closer to $1 trillion. A CNBC source called $1.5 trillion a “stretch,” and another said $1 trillion to $1.25 trillion looked more likely. The Financial Times, in a November 7 report recapped by other outlets, described a range of about $1.25 trillion to $1.75 trillion and a raise of $80 billion to $100 billion. Anthropic declined to comment to CNBC. None of this is company guidance.

This is a listing-and-valuation story about what investors are saying into an expected IPO window. It is not a power-equipment story, and it is not a research story. It is distinct from this morning’s report that Anthropic agreed to pay about $3.4 billion for Caterpillar gas-generator commitments via OnePWR, from METR’s assessment of Anthropic’s AI research-and-development pace, and from Anthropic’s November 6 research post on agent memory and “dreaming”. Those pieces cover generators, a third-party R&D-acceleration study, and an offline memory-cleanup design. They help explain why the company is raising so much capital and contracting for so much compute. They do not set the IPO price.

The calendar investors are watching

The timing is specific enough to put on a calendar and still loose enough that it could move. In a November 6 report by Corrie Driebusch, Berber Jin, and Anissa Gardizy, the Wall Street Journal (as carried by LiveMint) wrote that Anthropic is aiming to stage the IPO the week before Thanksgiving and has been expected to raise up to $100 billion in an offering that could value the company at around $1.5 trillion. U.S. Thanksgiving in 2026 falls on November 26, so “the week before” points at the trading week of November 16.

CNBC’s November 6 story, by Kate Rooney and Caleigh Keating, said Anthropic is set to list on the Nasdaq under the ticker ANT and is gearing up for an expected debut in a matter of weeks. The same piece said the company was expected to kick off its roadshow the following week — that is, the week of November 9, if that expectation held. A November 7 CNBC follow-up by Ashley Capoot described Anthropic as likely targeting mid-November for its Nasdaq debut.

A roadshow is the marketing tour in which a company’s executives and underwriters meet prospective buyers before shares are priced. “Expected to kick off” is not the same as “has begun,” and a target window is not a priced deal. The Journal also reported that Anthropic executives and its advisers were expected, as of November 6, to meet in the next few days to discuss IPO plans, and that they have avoided committing to specific plans throughout the process so they can keep their options open. We have not seen a public readout of that meeting.

How the valuation conversation moved

The investor discussion has been sliding down a set of unofficial rungs. They are worth keeping separate, because headlines tend to flatten them into a single “Anthropic is worth X” sentence.

Earlier talk, around $2 trillion. CNBC wrote that Anthropic had been looking toward a $2 trillion valuation in its IPO until recently, and that $2 trillion was the target when the company filed confidentially with the SEC. The same report said that last month Anthropic was reportedly seeking a valuation as high as $2 trillion, and that as of May the company was privately valued at $965 billion.
The more recent working assumption, around $1.5 trillion. The Journal described an offering that could value the company at around $1.5 trillion and raise up to about $100 billion. That is the number investors are now arguing about, not a price Anthropic has formally set in a public filing we have seen.
Pushback below $1.5 trillion. Three major investors who plan to participate told the Journal the valuation will likely need to come down, with one saying it could be closer to $1 trillion. CNBC reported that even $1.5 trillion is now in doubt, that one investor called that number a “stretch,” and that another said $1 trillion to $1.25 trillion is a more likely valuation target, although that could change with market conditions. The investors asked not to be named because they are not authorized to speak about the matter.
A wider range in the Financial Times. On November 7 the FT, as recapped by The Decoder and other secondary outlets, reported that Anthropic wants to go public before Thanksgiving and raise $80 billion to $100 billion at a valuation of about $1.25 trillion to $1.75 trillion. That range sits on both sides of the $1.5 trillion figure and above the lowest investor marks in the Journal and CNBC stories. It is still investor- and source-level reporting, not a published price range on an amended prospectus cover.

One way to read the spread is that the market has not agreed on a number. Another is that bankers often start a conversation above where they expect to print, then use the roadshow to find a clearing price. Neither reading is something Anthropic has endorsed. CNBC noted that Anthropic declined to comment.

Investors will expect the IPO to be priced at least at the low end of previous expectations. The market really needs this IPO to go well and the bankers on the deal surely understand that. The best way to set Anthropic up for success is to price the deal well.
Gil Luria, analyst at D.A. Davidson, speaking to CNBC, November 6, 2026

Luria’s comment is analyst color, not a bid. It does underline the unusual position this listing occupies: if the deal is as large as the reports suggest, the print will be read as a referendum on the AI trade, not only as a financing for one company.

What investors say could give — price, size, or timing

The Journal’s sources did not describe a single, agreed fallback. They described a menu. One investor said that even if the market bounces back, it could be a challenge to convince institutions that have just been burned in a selloff to take a big risk on the AI sector again, and wondered whether Anthropic could delay the offering. Another option, the paper wrote, is that the company could raise less money than anticipated.

That second option sits next to a cash figure the Journal published: Anthropic had about $146.6 billion in cash on hand as of September 30. A company holding that much cash is not in the position of a typical startup that must list in order to make payroll. It can, in principle, take a smaller primary raise, wait, or both. Whether it will do either is exactly the sort of decision the paper said executives and advisers had not locked in publicly.

The Journal also noted that investors had been eagerly anticipating the offering for months, with many calling Anthropic a “must-own” name even as other AI-related stocks wobbled. That enthusiasm and the newer skepticism can both be true at once: a name can be widely wanted and still be argued over at $1.5 trillion versus $1 trillion. Wanting the stock is not the same as wanting it at every price.

What the amended S-1 already shows — and what it does not

Anthropic’s amended Form S-1 is public, and it is the right place to look for contracted obligations, risk factors, and historical financials. It is not, in the coverage we have reviewed, the place where a final IPO price or a final offering size has been printed.

CNBC’s November 7 story noted that Anthropic said in an amended IPO prospectus that it has entered into purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion, “substantially all of which is related to compute and infrastructure.” We have covered that figure before in the context of the company’s build-out. It is a multi-year contracted-spend number, not a statement that Anthropic will write a check for $895.4 billion this quarter, and it is not an itemized list that names this morning’s reported generator transaction.

The OnePWR report and the prospectus figure should stay in separate columns. The Financial Times said Anthropic agreed in late September to pay about $3.4 billion for 1.7 gigawatts of Caterpillar gas-fired generator commitments via OnePWR Solutions, without acquiring the supplier. Reuters could not immediately verify that report, and both firms declined to comment. Even if the generator story is accurate, it does not tell you whether those commitments sit inside the $895.4 billion total, and it does not tell you what price public investors will pay for Anthropic’s equity. Power equipment is part of the reason the company needs a balance sheet this large. It is not a valuation model.

The same separation applies to METR’s November 5 report on AI R&D acceleration inside Anthropic, and to the lab’s November 6 research post on memory and dreaming. Those documents speak to how the company builds models and agents. An IPO prospectus and a roadshow speak to how public-market buyers will be asked to underwrite the cost of doing that at scale.

The market backdrop, as the papers reported it

The valuation pushback did not appear in a vacuum. CNBC and the Journal both placed it against a sharp week for AI-linked stocks and against political headlines after the U.S. midterm elections. The facts below are what those outlets reported. They are not an argument about how readers should vote or about what Congress will pass.

CNBC wrote that the Nasdaq plummeted more than 5 percent on Thursday, November 5, its steepest drop in 19 months, and that October had been the index’s worst month since March 2025, led by chipmakers and neoclouds. The Journal wrote that companies in the Nasdaq composite shed $2.5 trillion of value that Thursday. CNBC’s November 7 follow-up said voters opposed data-center-related ballot measures in 40 out of 41 cases, and cited an NBC News exit poll in which more than half of voters called AI data centers a “crucial” or “major factor” in their congressional choice.

CNBC also reported that Anthropic’s prospectus risk factors already flag organized opposition to data-center development, and that the company’s public profile could draw particular scrutiny to projects associated with it. That is a disclosure about risk, not a prediction that any named project will be blocked. It is still not a substitute for a priced deal.

OpenAI’s overlapping private raise belongs in the same backdrop paragraph, not in Anthropic’s term sheet. The Journal noted that OpenAI has raised less than half of the $30 billion it had initially targeted, and that it had so far received $11 billion at a $1.25 trillion valuation. That matches the financing story we cover separately in OpenAI’s reported $11 billion first close. Two large AI labs seeking fresh capital in the same fortnight is a market fact. It does not mean one company’s leaked private mark is a floor or a ceiling for the other.

What this is — and what it is not

Claim	Status	Notes
Anthropic aims to IPO the week before Thanksgiving (U.S. holiday Nov 26, 2026)	Reported investor/timing talk	Wall Street Journal, Nov 6; not a dated company invitation
Nasdaq ticker ANT; debut expected in a matter of weeks	Reported by CNBC	Nov 6: roadshow expected the following week; Nov 7: mid-November target
Earlier valuation talk around $2 trillion	Reported	CNBC: target when the company filed confidentially; last month as high as $2T
More recent discussion around $1.5 trillion, raise up to ~$100 billion	Reported working assumption	WSJ; not a final advertised range from the company
Major investors say $1.5 trillion may need to come down; one said closer to $1 trillion	Investor talk to the Journal	Three major investors who plan to participate; unnamed
One CNBC investor called $1.5 trillion a “stretch”; another said $1–1.25 trillion is more likely	Investor talk to CNBC, Nov 6	Unnamed; “could change with market conditions”
FT range about $1.25–1.75 trillion, raise $80–100 billion	Reported Nov 7, via secondary recaps	Still not a published cover price
Cash about $146.6 billion as of Sept 30	Reported by the Journal	Cited in the context of possibly raising less than planned
Amended S-1 discloses about $895.4 billion of compute/infrastructure commitments	In the prospectus, as quoted by CNBC and prior coverage	Multi-year contracted spend; not an IPO price
This is the same story as the OnePWR generator report, METR’s R&D study, or dreaming/memory research	No	Those are infrastructure and research notes
Anthropic has commented on the valuation debate	No	CNBC: Anthropic declined to comment
A delay or a smaller raise has been decided	Not established	Journal sources floated both as options; no public decision

FAQ

Has Anthropic set a final IPO valuation?

No. The figures in circulation — $2 trillion earlier, about $1.5 trillion more recently, investor marks nearer $1 trillion to $1.25 trillion, and an FT range of about $1.25 trillion to $1.75 trillion — come from people speaking to newspapers and television, not from a priced deal. Anthropic declined to comment to CNBC.

When is the IPO expected?

The Journal said Anthropic is aiming for the week before Thanksgiving. In 2026 that holiday is Thursday, November 26, so the implied window is the week of November 16. CNBC has described a mid-November Nasdaq debut and, as of November 6, a roadshow expected to start the following week. Windows move. Nothing in those reports is a priced date.

What ticker would Anthropic use?

CNBC reported that Anthropic is set to list on the Nasdaq under the ticker ANT. That is reporting about the planned symbol, not evidence that shares are already trading.

How much cash does Anthropic have?

The Journal reported about $146.6 billion in cash on hand as of September 30, 2026. The paper raised that figure in the context of a possible decision to raise less money than the previously discussed offering size.

What is the $895.4 billion number?

It is the aggregate committed spend on purchase commitments and lease agreements disclosed in Anthropic’s amended IPO prospectus, substantially all related to compute and infrastructure. It is not the IPO valuation, not the size of the offering, and not a confirmation of the reported $3.4 billion OnePWR generator transaction.

Is this the same story as this morning’s generator report, or as METR and dreaming?

No. This morning’s piece is about a reported take-over of Caterpillar generator commitments via OnePWR. The METR article is about a third-party study of Anthropic’s research pace. The dreaming article is about how Claude reorganizes stored memory between sessions. This article is about what investors are saying about the company’s expected listing price and timing.

Could Anthropic delay the IPO or raise less?

Both were raised as possibilities by people who spoke to the Journal. One investor wondered about a delay; the paper separately noted that the company could raise less than anticipated. Anthropic has not announced either decision.

Did Anthropic comment?

CNBC reported that Anthropic declined to comment. Treat the valuation marks as investor conversation until the company or its underwriters publish a price range and then a final print.

Sources: Corrie Driebusch, Berber Jin, and Anissa Gardizy, The Wall Street Journal, via LiveMint (November 6, 2026); Kate Rooney and Caleigh Keating, CNBC, November 6, 2026; Ashley Capoot, CNBC, November 7, 2026; The Decoder recap of the Financial Times, November 7, 2026. The Journal’s cash figure, Thanksgiving-week aim, $1.5 trillion / up-to-$100 billion working assumption, and investor remarks about a lower print or a possible delay come from that November 6 report. CNBC is the source for the ANT ticker, the “stretch” quote, the $1–1.25 trillion investor mark, the expected roadshow timing, Luria’s comments, the declined-comment line, and the prospectus language on $895.4 billion of commitments. We have not seen a priced offering document, and we are not treating unnamed-investor marks as company guidance.

November 9, 2026

OpenAI Closes About $11 Billion at $1.25 Trillion Valuation, Bloomberg Reports (2026)

Takeaway: On November 6, 2026, Bloomberg reported that OpenAI had completed an initial close of about $11 billion in a new funding round that values the company at $1.25 trillion. People familiar with the matter told the outlet the money had been wired by Friday, and that the rest of the round could still take the total to as much as about $20 billion in the coming weeks. Reuters separately reported the same $11 billion initial close and $1.25 trillion valuation, citing a person familiar with the matter. OpenAI has not published its own confirmation of the wire. Recaps of Bloomberg’s account say the company declined to comment; TechCrunch said it had reached out and would update if OpenAI responded. The same day, OpenAI chief financial officer Sarah Friar told CNBC’s “Squawk on the Street” that the company is in a current round, would raise at a $1.25 trillion pre-money valuation, and that the round is ongoing.

This is a financing story about how much private capital OpenAI has taken in, at what price, and what its CFO said about the business while that round was still open. It is not a product launch and should not be collapsed into this week’s notes on the Pierre Computer acquisition, GPT-6 Cyber, the Hacktron defensive-testing partnership, or OpenAI’s Exponential Time Hypothesis research claim. Those pieces cover an acquisition, a cybersecurity model, a security engagement, and a math claim still under outside review. The dollars below do not change any of those facts.

What Bloomberg said was wired — and what is still open

The core of the November 6 report is narrower than a completed mega-round. Bloomberg’s sources described about $11 billion as the first portion of a larger raise, already transferred to the company by Friday. One of those people said the remaining amount is expected over the coming weeks and could bring the total to as much as about $20 billion. Secondary write-ups of that Bloomberg story (including Storyboard18 and LiveMint) repeat those figures and add that OpenAI had spent more than a month discussing a fresh capital raise after pushing back its own public-listing plans.

Two distinctions matter for anyone tracking the headline. First, a first close is not the same thing as a finished round. Friar said on CNBC that the round is ongoing. Bloomberg’s sources said more money could still come in. Until OpenAI or a filing says the round is done, treat $11 billion as the reported cash already in, not as a final total.

Second, a commitment and a wire are not the same sentence. In late October, the Wall Street Journal reported that OpenAI had already secured about $11 billion in commitments at a $1.25 trillion valuation, less than half of a $30 billion aim, and expected to raise more in the coming weeks. TechCrunch recapped that Journal report on October 31 as a follow-up to a Financial Times story the previous day. If both accounts are right, the November 6 news is that money previously described as committed has now been described as wired — not that an entirely new $11 billion suddenly appeared.

None of these figures sits in a document OpenAI has published. There is no press release, no shareholder letter, and no securities filing that we have seen which itemizes the close. A report from unnamed sources is the news. It is not an audited capital-account statement.

How the target shrank, according to the papers

The $11 billion first close, and the talk of a round that might still reach about $20 billion, sit well below what outlets had been writing earlier in the autumn. Bloomberg previously reported that OpenAI was looking to raise about $30 billion or more at a valuation around $1.4 trillion. TechCrunch’s November 6 story recalled a late-September Bloomberg report that put the aim at $30 billion.

On October 30, the Financial Times reported that OpenAI had subsequently lowered its funding target to about $20 billion, with a first close expected to come in even lower than that, according to TechCrunch’s recap. The FT said a group of investors from the United Arab Emirates would lead the round, and that a later offering aimed at wealthy individuals could help the company reach the $20 billion figure. The Journal’s late-October follow-up, as recapped the next day, described investor response as tepid and put commitments already in hand at about $11 billion.

The papers are not in perfect agreement about the earlier valuation hope. Bloomberg’s run of stories uses about $1.4 trillion. A Reuters account of the November 6 close, citing two investors, said the company’s valuation had come down from a $1.5 trillion level it had initially sought, against an original fundraising target of up to $30 billion. Those are not the same starting number. Treat the original target as something the papers themselves do not fully settle, and treat every one of these figures as sourced reporting rather than a number OpenAI has posted on its own site.

The Verge, summarizing Bloomberg on November 6, tied the smaller raise and lower headline valuation to a drop in tech stocks and to midterm-related opposition to data-center projects. That is context, not proof of why any single investor wrote or withheld a check.

The March round this price is being compared with

OpenAI’s previous major private raise, in March 2026, brought in about $122 billion at a valuation of $852 billion, including that raise, according to The Verge and other recaps. After Eisen noted reports of a lower valuation and a smaller raise than originally expected, Friar said the company is in a current round, would raise at a $1.25 trillion pre-money valuation, and that the round is ongoing but that OpenAI feels quite confident in it.

I think the original, like, started with news from another outlet talking about 1.2. In the end, 1.25, remember, is almost a 50 percent lift in valuation from the round we closed in March. So I genuinely think that that is a good indication of our Q3 momentum, and, in particular, how that momentum carries into 2027.
Sarah Friar, OpenAI CFO, on CNBC’s “Squawk on the Street,” November 6, 2026

If the March valuation was $852 billion, $1.25 trillion is roughly a 47 percent increase, matching Friar’s “almost a 50 percent lift.” That comparison does not settle whether $1.25 trillion is generous or disappointing relative to the $1.4 trillion (or, in some Reuters sourcing, $1.5 trillion) earlier stories attached to the company. “Pre-money” is the price before new cash is added; a later full close near $20 billion would imply a higher post-money number. Until OpenAI publishes term-sheet details, do not treat those two figures as interchangeable, and do not assume the extra $9 billion has already arrived.

What Friar said about the business, not just the round

The CNBC interview is the on-the-record window we have from inside OpenAI the same day as the wiring report. Asked about fundraising after a week of headlines about data-center opposition following the U.S. midterms, Friar said the demand for AI is real, that her job is to manage through scenarios, and that OpenAI would keep raising capital. She did not confirm Bloomberg’s $11 billion wire figure and was not asked to name investors.

She did give several business figures that help explain why a private round at this scale is even being discussed:

The company’s business “grew 70 percent quarter over quarter” in what she called a breakout third quarter.
The enterprise business “doubled” in the same quarter. Later in the interview she said the company is now “a very balanced business, 50 percent from consumer, 50 percent from enterprises.”
“1.2 billion people across the globe come to us every single week,” she said, including 300 million for health and wellness and 200 million for personal finance. Those are her figures, not an independent measurement.

Asked about a public listing, Friar again said a good CFO raises money when she can. One reason to go public, she said, beyond capital OpenAI has already raised in private markets, is to put economic value “into as many hands as we can.” When Eisen asked whether that is a 2027 story, Friar said, “I think last we said we would continue to look in ’27.” That restates prior guidance; it is not a new filing date. Asked whether Anthropic listing first complicates OpenAI’s raise, she said OpenAI should “run your own race.” She had looked at Anthropic’s S-1, she said, but she did not comment on Anthropic’s price.

Which investors have been named — and by whom

OpenAI has not published a cap table for this close. The names in circulation come from people speaking to reporters.

Bloomberg recaps say Coatue Management and T. Rowe Price were among the investors in the first close, and that SoftBank Group, Sequoia Capital, Founders Fund, and Andreessen Horowitz are among the investors in the broader round. Those recaps also say a group of investors from the United Arab Emirates has joined, and that OpenAI is working with investment banks to raise money from high-net-worth individuals and retail investors.
Reuters, citing investors, said SoftBank and MGX participated in the funding. MGX is an Abu Dhabi-based investor; that Reuters line may overlap with Bloomberg’s description of a UAE investor group, but the two accounts should not be flattened into a single, company-confirmed lead.
The October 30 Financial Times report, as recapped by TechCrunch, had already said a UAE investor group would lead the then-revised round. That is consistent in direction with the later Bloomberg and Reuters mentions of Gulf participation. It is still not a primary confirmation from OpenAI.

Named by a source is not the same as named by the company. Ticket sizes, any mix of primary and secondary shares, and special rights attached to the new stock have not been disclosed in the coverage we reviewed.

How this sits next to Anthropic’s listing talk

The OpenAI raise landed in the same week investors were arguing about the price of Anthropic’s planned IPO. That is useful market context, and it is a different story. Our separate note on valuation pushback around Anthropic’s Thanksgiving-week listing window covers that debate. OpenAI is still private and, if Bloomberg is right, is taking in a first slice of a round smaller than the $30 billion-class figure floated in September. Anthropic has an amended public S-1 and, according to the same cluster of reports, is trying to list before Thanksgiving. Do not use one company’s leaked private valuation as a mark-to-market for the other.

The infrastructure backdrop is shared even if the securities are not. Anthropic’s prospectus discloses about $895.4 billion of compute and infrastructure commitments, a figure we discussed alongside this morning’s reported $3.4 billion gas-generator commitments via OnePWR and METR’s write-up of Anthropic’s AI research-and-development pace. Those notes help explain why AI labs are raising and listing at these sizes. They are not evidence about OpenAI’s $11 billion wire.

What this is — and what it is not

Claim	Status	Notes
OpenAI closed about $11 billion as the first part of a round valuing it at $1.25 trillion	Reported, not company-confirmed	Bloomberg, Nov 6, citing people familiar; Reuters cited a person familiar with the same close
The $11 billion had been wired by Friday, November 6	Reported by Bloomberg	OpenAI declined to comment, according to recaps of that story
The rest of the round could bring the total to as much as about $20 billion	Reported expectation	Bloomberg sources; Friar said the round is ongoing
The $1.25 trillion figure is pre-money	Stated by Friar on CNBC; used in Bloomberg recaps	Pre-money is before new cash; a later full close would imply a higher post-money number
OpenAI originally sought about $30 billion at about $1.4 trillion	Earlier Bloomberg reporting	Some Reuters sourcing of the Nov 6 close instead cites an earlier $1.5 trillion hope
The FT said in late October that the target had been cut to about $20 billion	Reported	TechCrunch recap of the Oct 30 FT story; WSJ then reported ~$11 billion already committed
March round: about $122 billion at an $852 billion valuation	Widely reported, including by The Verge	Friar called 1.25 almost a 50% lift vs that March round
Q3 business +70% quarter over quarter; enterprise doubled; ~1.2 billion weekly ChatGPT users	Stated by Friar on CNBC, Nov 6	Company figures; we did not independently audit them
IPO is a 2027 story	Prior guidance restated	Friar: “I think last we said we would continue to look in ’27.”
Coatue and T. Rowe Price in the first close; SoftBank, Sequoia, Founders Fund, a16z in the round; Reuters: SoftBank and MGX	Named by sources, not by OpenAI	Bloomberg recaps vs Reuters investors; names may overlap and are not a cap table
This is the same story as Pierre, GPT-6 Cyber, Hacktron, or the ETH claim	No	Those are product, security, and research notes

FAQ

Did OpenAI confirm that $11 billion was wired?

No. The wire is a Bloomberg report, citing unnamed people familiar with the matter, later echoed by Reuters. Recaps of Bloomberg’s story say OpenAI declined to comment. Friar’s CNBC interview the same day confirmed that a round is underway at a $1.25 trillion pre-money valuation and that it is still open. She did not confirm the $11 billion figure.

Is $11 billion the whole round?

Not according to the reporting. Bloomberg’s sources described it as an initial close, with more money possibly taking the total to as much as about $20 billion in the coming weeks. Friar said the round is ongoing. The late-October Financial Times story had already described a reduced target of about $20 billion.

What does a $1.25 trillion pre-money valuation mean?

It is the price of the company used to sell the new shares, counted before the new cash is added. If investors later complete a larger close, the post-money valuation would usually be the pre-money figure plus the primary proceeds. OpenAI has not published the share count or the split between primary and secondary stock.

How does this compare with OpenAI’s March round?

The March 2026 round was reported at about $122 billion raised and an $852 billion valuation, including that raise. Friar told CNBC that 1.25 is almost a 50 percent lift from that March valuation. Relative to earlier autumn stories about a $1.4 trillion (or, in some Reuters sourcing, $1.5 trillion) target, the current headline number is lower.

Who is investing?

OpenAI has not released an official list. Bloomberg recaps name Coatue Management and T. Rowe Price among first-close participants, and SoftBank, Sequoia, Founders Fund, and Andreessen Horowitz among investors in the broader round, plus a UAE investor group. Reuters, citing investors, named SoftBank and MGX. Treat those as sourced names, not a company cap table.

When will OpenAI go public?

Friar told CNBC that the company would continue to look in 2027, repeating earlier guidance. She did not name a quarter. This funding story is about a private round, not an S-1.

Is this related to GPT-6 Cyber, Hacktron, Pierre, or the ETH paper?

Only in the weak sense that the same company is involved. The financing does not change the terms of the Pierre acquisition, the GPT-6 Cyber launch, the Hacktron testing partnership, or the still-unverified Exponential Time Hypothesis claim. Those remain separate stories.

Sources: Bloomberg — OpenAI Completes $11 Billion Portion of Funding at $1.25 Trillion Value (November 6, 2026); Reuters, via Investing.com; Reuters, via CNBC-TV18 (SoftBank and MGX); TechCrunch, November 6; TechCrunch, October 31 (FT target cut and WSJ commitments recap); The Verge; Storyboard18 and LiveMint recaps of Bloomberg’s investor names and $20 billion remainder; CNBC unofficial transcript of Sarah Friar on “Squawk on the Street,” November 6, 2026. Quotations and business figures attributed to Friar are taken from that CNBC transcript. We have not seen an OpenAI press release or filing that confirms the wire, the investor list, or the final size of the round.

November 9, 2026

Anthropic Agrees to Pay About $3.4 Billion for Gas Generators, FT Reports (2026)

Takeaway: On November 8, 2026, the Financial Times reported that Anthropic agreed in late September to pay about $3.4 billion for commitments covering 1.7 gigawatts of heavy-duty, Caterpillar gas-fired generators. According to the paper, the AI lab took over existing orders placed by OnePWR Solutions, a San Antonio, Texas equipment supplier, along with engineering and construction capability needed to install the machines. Anthropic did not acquire OnePWR itself, and no OnePWR employees transferred to the startup, the FT said, citing a person familiar with the matter. The newspaper framed the transaction as the largest known single purchase of power-generation equipment by an AI laboratory. CNBC, Bloomberg, and other outlets followed the same day. Reuters said it could not immediately verify the report, and both Anthropic and OnePWR declined to comment to the FT and did not respond to Reuters’ requests.

This is a power-equipment story about how an AI lab is trying to get around generator lead times by taking over someone else’s Caterpillar queue, rather than waiting in line for new factory slots. It is not a corporate acquisition, not a disclosed site announcement, and not a statement of where the machines will run — the FT said that location is not known. It is distinct from Google’s November 5 plan for a $17 billion Wyoming data-center campus paired with a 702 MW fuel-cell block, and from OpenAI’s Grid Protectors program, which is about treating AI campuses as flexible grid load rather than buying a fleet of gas generators outright.

What the Financial Times said Anthropic bought

The core claim is simple, and it is still a report rather than a company disclosure. In late September 2026, Anthropic signed an agreement worth about $3.4 billion to secure 1.7 gigawatts of large gas-fired generators that OnePWR had already ordered from Caterpillar. Follow-up coverage in The Next Web and elsewhere, attributing the same FT account, added that OnePWR placed those Caterpillar orders about two years earlier and needed financing to complete the purchase. Anthropic’s first idea, the paper said, was to buy electricity from OnePWR. Instead, it stepped into the equipment orders themselves and took the installation work that goes with them.

If the reported totals are accurate, $3.4 billion for 1.7 gigawatts works out to roughly $2 million per megawatt of committed generation. That is a back-of-the-envelope figure from the two numbers the FT published, not a quoted unit price, and it would cover whatever mix of machines, transportation, and install capability was bundled into the commitments. The reports do not name Caterpillar model numbers, do not say how many individual generator sets are involved, and do not say whether the $3.4 billion is payable up front or over the life of the orders.

1.7 gigawatts is a large block of on-site or near-site thermal capacity by data-center standards. For scale, Google’s Wyoming announcement last week highlighted a contracted 702 MW fuel-cell development — less than half the generator capacity the FT attaches to Anthropic — and even that figure described generation next to a campus, not a published IT load. One and seven-tenths gigawatts is also in the same neighborhood as the compute capacity Anthropic has been telling markets it expects to have available. Data Center Dynamics, summarizing Anthropic’s amended IPO prospectus and a New York Times figure, wrote that the company is expected to have more than 3.5 GW of compute this year, with a similar amount due in 2027. A 1.7 GW generator commitment would not, by itself, power that entire compute plan, and no public document we have seen maps these machines onto a named campus.

What Anthropic did not buy

The structure matters as much as the price. The FT is explicit that Anthropic did not acquire OnePWR Solutions and that none of the supplier’s staff moved across. What transferred, according to the person the paper cited, were the generator commitments plus the engineering and construction capabilities needed to install them. In other words, this is described as a take-over of orders and know-how around a piece of kit, not a vertical merger that would put a Texas equipment firm on Anthropic’s books.

That distinction is easy to lose in headlines that say an AI company “bought generators” or “bought a power supplier.” Buying the metal and the right to finish installing it is different from buying the company that originated the purchase. It is also different from a long-term power-purchase agreement in which OnePWR would have owned the generators and sold Anthropic electrons. The reported path — plan to buy power, then take over the equipment orders when the supplier needed money to complete them — is a reminder that AI infrastructure is now colliding with ordinary industrial-finance constraints: factory slots, deposits, and the cost of carrying heavy equipment before a data hall is ready.

Why a two-year-old Caterpillar queue is valuable

Large reciprocating engines and packaged generator sets are not software. Caterpillar and its peers have been quoting long lead times as data-center and grid customers compete for the same production lines. An order placed roughly two years ago, if the FT’s account is right, sits ahead of anyone walking into a dealer in late 2026. Taking over that queue is a way to buy time rather than just capacity. The install capability bundled into the deal points the same direction: owning a generator that cannot be placed, fueled, permitted, and interconnected is not the same as having megawatts.

On-site gas generation has become a common bridge for AI campuses that cannot wait for new transmission or a utility-scale plant. Google’s Wyoming package pairs fuel cells and carbon capture with power from gas generators operated by partners. Other operators have used temporary turbines to energize halls while longer-term interconnection proceeds — and have drawn legal and community pushback for it. The Next Web, in its write-up of the Anthropic report, noted that Elon Musk’s xAI faces a lawsuit over gas turbines at its Memphis site. That is context for the sector, not evidence about Anthropic’s own permitting situation, which has not been disclosed in connection with this deal.

Gas generators also come with fuel supply, air permits, noise, and — if they run as more than emergency backup — a carbon profile that sits awkwardly next to corporate climate pledges. None of the November 8 stories say how many hours a year Anthropic intends to run these machines, whether they are meant as primary power or as a bridge, or whether carbon capture is part of the plan. Those are open questions, not implied answers.

What has been independently corroborated — and what has not

Readers should keep the verification stack in view. The original reporting is the Financial Times on Sunday, November 8, based on a person familiar with the matter. Both companies declined to comment to the FT. Bloomberg followed the same report; The Next Web said the companies also did not immediately respond to Bloomberg. CNBC’s write-up, published that morning in the U.S., stated plainly that Reuters could not immediately verify the report and that Anthropic and OnePWR did not respond to its requests for comment. We have not seen a press release, an SEC exhibit, or a signed contract.

There is one adjacent public breadcrumb that does not name Anthropic. The Energy Mag reported that a separate transaction disclosure by Piper Sandler lists a $3.4 billion sale of OnePWR’s power-generation platform to an unnamed AI lab in September. Piper Sandler, which the same article says served as OnePWR’s exclusive financial adviser, does not identify the buyer as Anthropic. A $3.4 billion September sale of that platform is consistent with the FT’s timing and price, but it is not the same thing as a named confirmation. Until Anthropic, OnePWR, Caterpillar, or a regulatory filing says so, the buyer’s identity remains a reported claim.

Reuters could not immediately verify the report and Anthropic and OnePWR did not respond to requests for comment.
CNBC, summarizing the Financial Times report, November 8, 2026

How this sits next to Anthropic’s IPO paperwork

The reported generator purchase landed as Anthropic is preparing to list. In an amended Form S-1 filed around November 2, 2026, the company disclosed purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion, “substantially all of which is related to compute and infrastructure,” according to Data Center Dynamics’ reading of the prospectus. That figure had previously been listed around $417 billion for compute.

The S-1 math is about contracted compute and infrastructure over many years, not about this generator order specifically. Public recaps of the filing break out large commitments to Google, SpaceX, Amazon, Microsoft, and a set of neoclouds; they do not, in the coverage we reviewed, itemize a Caterpillar or OnePWR line. It would be a mistake to add $3.4 billion to $895.4 billion or to assume the generator commitments are already inside that total. Anthropic has not said. What the prospectus does show is the scale of the bet: the company is contracting for computing capacity measured in hundreds of billions of dollars and, per the Times figure relayed by DCD, multi-gigawatt compute availability over 2026 and 2027. Power equipment is the unglamorous prerequisite for that compute actually turning on.

A separate Reuters report in September said Anthropic expects to spend more than $500 billion over the next ten years building AI infrastructure. That longer-horizon spending figure and the S-1’s $895.4 billion of contractual commitments are not the same metric — one is an expectation about outlays, the other is a tally of signed purchase and lease obligations — and they should not be averaged together. Both point in the same direction: Anthropic is trying to lock down scarce inputs years before models trained on them ship.

Anthropic’s earlier energy pledges

Buying gas generators also sits next to policy Anthropic has already put on the record. The Energy Mag, in its follow-up to the FT story, recalled a February announcement in which Anthropic pledged to pay for grid upgrades needed to connect its data centers, to bring new generation online to match their electricity needs, and to cover consumer electricity-price increases attributable to those facilities. The same account said those commitments apply directly where Anthropic works with partners to develop data centers for its own workloads, and that the company was still exploring how to address price effects when it leases capacity in existing facilities.

What operators should actually take from this

For people who watch AI capacity, the useful signal is not that one lab “spent $3.4 billion on energy.” It is that factory queues for heavy generation gear have become a strategic asset, valuable enough that an AI company would rather assume someone else’s Caterpillar orders than wait. Lead time, not nameplate efficiency, appears to be the scarce resource. That differs from Google’s Wyoming package (generation contracted next to a named campus) and from OpenAI’s Grid Protectors work (flexible load).

For people who run or site data centers, the reported structure is a case study in how offtakers are moving upstream. Anthropic is described as stepping into equipment procurement and installation, not merely signing a power-purchase agreement. That can shorten the path to electrons. It also puts more of the permitting, interconnection, fuel, and community burden onto the lab — or onto whatever partners it later hires to place the machines. Because the FT says it is not known where the generators will be used, neighboring communities cannot yet evaluate the local impact, and outsiders cannot tell whether this capacity is tied to a new campus, to several sites, or to a mobile/temporary deployment model.

This coverage is also not a recap of Anthropic’s other news this month. The company’s research transparency work with METR on AI R&D acceleration and its notes on Claude agent memory and “dreaming” are separate stories about models and evaluation. The generator report is about the physical layer underneath those models.

What this is — and what it is not

Claim	Status	Notes
Anthropic agreed in late September 2026 to pay about $3.4 billion for 1.7 GW of gas-fired generator commitments	Reported, not independently confirmed	Financial Times, Nov. 8, 2026; person familiar with the matter
The equipment was ordered from Caterpillar by OnePWR Solutions of San Antonio	Reported by the FT	Orders described as about two years old; OnePWR needed financing to complete them
Anthropic took over the orders plus install/engineering capability	Reported by the FT	Anthropic first planned to buy power from OnePWR, then took the equipment path
Anthropic acquired OnePWR or hired its staff	No, per the FT	Explicitly not an acquisition; no employee transfer
This is the largest known single power-equipment purchase by an AI lab	FT framing	A characterization by the newspaper, not a ranking Anthropic published
Where the generators will be installed	Unknown	The FT said the location is not known
Reuters independently verified the deal	No	CNBC: Reuters could not immediately verify; both firms declined or did not comment
Piper Sandler disclosed a $3.4 billion September sale of OnePWR’s platform to an AI lab	Reported as a public disclosure that does not name Anthropic	Consistent on price and month; buyer unnamed in that disclosure
Anthropic’s amended S-1 lists about $895.4 billion of compute and infrastructure commitments	Confirmed in coverage of the Nov. 2, 2026 filing	Not itemized as this generator order
The $3.4 billion is included in, or extra to, the $895.4 billion	Not stated	Do not add the figures together

FAQ

Did Anthropic buy OnePWR Solutions?

Not according to the Financial Times. The paper reported that Anthropic took over OnePWR’s Caterpillar generator orders and related installation capability but did not acquire the San Antonio supplier and did not take on its employees.

Is the $3.4 billion deal confirmed?

It is confirmed only as an FT report dated November 8, 2026. CNBC noted that Reuters could not immediately verify it. Anthropic and OnePWR declined to comment to the FT and did not respond to Reuters. A Piper Sandler disclosure of a $3.4 billion September sale of OnePWR’s power-generation platform to an unnamed AI lab is consistent on price and timing but does not name Anthropic.

Where will the 1.7 gigawatts of generators be used?

The FT said it is not known. None of the follow-up stories we reviewed named a state, a campus, or a utility interconnection. Treat any map that appears in secondary commentary as speculation unless a primary document appears.

Why would Anthropic take over someone else’s Caterpillar orders?

Because large generator sets have long factory lead times. Coverage of the FT story said OnePWR ordered the machines from Caterpillar about two years ago and needed money to finish the purchase. Stepping into that queue is a way to secure equipment that would take much longer to order new — a different tactic from contracting a utility or a fuel-cell partner to build generation at a named site.

How does this compare with Google’s Wyoming plan or OpenAI’s Grid Protectors program?

Google’s November 5 Wyoming announcement is a long-term campus investment that highlights a 702 MW fuel-cell development with carbon capture, plus partner-operated gas generation. OpenAI’s Grid Protectors work is about making AI load flexible for the grid. The Anthropic report is about taking over a large block of Caterpillar gas-generator commitments, without a named site. The three stories share a theme — AI labs are now energy actors — but they describe different tools.

Is this part of the $895.4 billion in Anthropic’s amended S-1?

Anthropic has not said. The amended prospectus describes about $895.4 billion of purchase commitments and lease agreements, substantially all for compute and infrastructure. Public recaps do not itemize a OnePWR or Caterpillar line. The $3.4 billion figure should not be added to or subtracted from that total without a company breakdown.

Will these generators run all the time?

No public account of the deal says so. Gas units can serve as primary power, as a bridge until grid or renewable capacity arrives, or as backup. Run hours, fuel contracts, air permits, and any carbon-capture plan have not been disclosed in the November 8 reporting.

Sources: CNBC — Anthropic secures $3.4 billion generator deal for AI data centers, FT reports (November 8, 2026); The Next Web — Anthropic buys $3.4bn of gas generators for its data centres, FT reports (November 8, 2026); The Energy Mag — Anthropic Agrees to $3.4 Billion Gas-Generator Deal, FT Reports; Data Center Dynamics — Anthropic S-1 update reveals $895.4bn in cloud and data center commitments (November 3, 2026). The underlying Financial Times article is paywalled; figures and characterizations above are taken from these secondary reports and are labeled as such. We have not seen the contract, a Caterpillar order book, or an Anthropic or OnePWR confirmation.

November 9, 2026