ProductIndependently confirmed2 publishers2 min readPublished Updated
Anthropic weighed a $7B MatX deal to pull chip design in house
Reuters reports the acquisition talk has cooled into a partnership, but the $7 billion figure and two senior chip hires say Anthropic treats silicon as a cost it has to own rather than a vendor it can pick.
The Product Desk
What happened
- Anthropic discussed buying AI chip start-up MatX for roughly $7 billion, two people briefed on the matter told Reuters, and a third said the talks have since become a discussion about a partnership.
- MatX, founded by former Google tensor processing unit engineers, is now seeking new capital at a valuation of about $4 billion, according to one of the people who spoke to Reuters.
- Anthropic has met a range of AI chip start-ups in recent weeks to understand current design approaches, and has not yet elected to make an acquisition.
- Its named compute spending includes $36 billion of Google AI chips, a $45 billion deal to rent cloud capacity from Nscale, and $1.25 billion a month to SpaceX through May 2029.
Why it matters
- cost Owning the design means owning hundreds of millions of dollars per generation and a wait of a year or more before working parts exist, so the spending lands long before any saving does.
- constraint With Nvidia telling investors its processors stay scarce through 2027, Anthropic's plan to buy chips directly is rationed by someone else's allocation until it has parts of its own.
- exposure A listing expected in 2026 puts a $2 trillion ask and a 2028 revenue number in front of investors years before in-house silicon could show up in gross margin, so they are underwriting intent.
- contradiction Anthropic's public position is an expanded silicon team and continued multi-vendor buying, which describes a far smaller commitment than the acquisition price its would-be counterparty was reportedly discussed at.
Whether this touches you depends on whether it's training or inference. MatX has been working on a chip for training large models [7], which sets what it costs Anthropic to build the next Claude rather than what it costs anyone to call the current one. OpenAI's claim at a conference this week that its Jalapeno part beat a comparable Nvidia processor rested on energy efficiency in inference calculations [16], the side of the line a buyer sees on a rate card. Anthropic's sources say it could elect to produce an inference chip as well [10].
The case for owning the design is a cost-of-goods case, and the people briefed on the talks put it plainly: buying a start-up like MatX would give Anthropic in-house design expertise and could help lower costs over the longer term [11]. Set that against what renting costs. The Google chip purchase and the Nscale deal alone come to $81 billion in named commitments [21], and the SpaceX arrangement annualises to roughly $15 billion a year [22], more than twice the price discussed for MatX [23]. That discussed price also sat about 75 percent above the valuation MatX is now asking investors for [20]. Seen from the rent side of the ledger, a design team is cheap to buy. Anthropic declined to comment on the deal talks and MatX did not respond [5], and Reuters could not learn why the talks are no longer active [3].
For the person who has to sign a commitment, two questions sort any lab's chip news: whether it's training or inference, and whether it's money committed or people committed. The chip purchases and cloud deals are money, with dates attached. The MatX outcome so far is people: a partnership discussion in place of a purchase, plus Google chip veteran Amir Salek hired this week and former OpenAI chip engineer Clive Chan hired in June [15]. Only the inference-and-money quadrant should move your assumptions about per-token pricing over the next two years. The people column tells you something slower and still worth knowing, which is where a vendor expects its margin to come from at the end of the decade.
The tradeoff in that rule is that it makes you late. Google builds TPUs and Amazon builds Trainium and Inferentia [19]. Custom silicon from a model lab is not a new idea, and if Anthropic's own silicon ever does for inference what OpenAI says Jalapeno does, customers will most likely notice it first in a quieter rate card, not a keynote.
What to watch
- Whether MatX closes a round near $4 billion, or either party describes the partnership in public.
- Whether Anthropic turns any of its recent chip start-up meetings into an actual acquisition.
- Whether Nvidia revises the 2027 supply picture that the in-house design effort is meant to hedge.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+20
- Incentives60
- Confidence50
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic discussed buying AI chip start-up MatX for roughly US$7 billion to accelerate its custom hardware efforts, according to two people briefed on the matter who spoke to Reuters.
ReportedSupportedSource: Reuters, via two people briefed on the matter2 sources— create a free account to open themView cited source - [2]
A third person said the merger talks have evolved into a discussion about a partnership.
ReportedSupportedSource: Reuters, via a third person2 sources— create a free account to open themView cited source - [3]
Reuters could not learn why the Anthropic-MatX talks are no longer active.
- [4]
MatX was founded by former Google tensor processing unit engineers and is now seeking to raise new capital at a valuation of about US$4 billion, one of the people said.
- [5]
Anthropic declined to comment on deal talks with MatX, and MatX did not respond to a request for comment.
- [6]
Anthropic said it is expanding an in-house silicon team to design custom chips that will let Claude models run faster and more efficiently, and plans to keep a multi-chip approach by working with providers including Nvidia and Google.
- [7]
MatX has been working on a chip that would be useful for building large AI models, a process called training.
- [8]
Anthropic plans to buy US$36 billion worth of Google's AI chips.
- [9]
Anthropic signed a US$45 billion deal to rent AI cloud computing power from Nscale.
- [10]
The MatX discussions suggest Anthropic may be interested in producing a training chip, while other chip start-ups and OpenAI pursue inference processors; sources said Anthropic could elect to produce an inference chip as well.
- [11]
An acquisition of an AI chip start-up such as MatX would give Anthropic in-house design expertise and could help lower costs over the longer term, the people said.
ReportedSupportedSource: people briefed on the matter, via Reuters2 sources— create a free account to open themView cited source - [12]
Designing chips is expensive and time-consuming: it can take a year or more to produce a viable piece of hardware, and design costs for a single generation run in the hundreds of millions of dollars.
- [13]
Anthropic agreed to pay SpaceX US$1.25 billion per month through May 2029 for computing capacity across its data centre clusters.
- [14]
Anthropic is expected to list in 2026, and Reuters reported earlier in August that the IPO will chase a US$2 trillion valuation hinging on a 2028 revenue figure of as much as US$200 billion.
- [15]
Anthropic hired Google chip veteran Amir Salek this week, and in June hired former OpenAI chip engineer Clive Chan, who worked on the OpenAI chip unveiled earlier in 2026.
- [16]
At a conference this week OpenAI said its first custom chip, called Jalapeno, outperformed a similar Nvidia processor, and executives noted it was more energy efficient in performing inference calculations.
- [17]
Nvidia said on an Aug 26 conference call that its processors would be in short supply through 2027.
- [18]
In recent weeks Anthropic held meetings with a range of AI chip start-ups to understand the current range of chip design approaches, and has not yet elected to make an acquisition.
- [19]
Google has developed TPUs, while Amazon has built its Trainium and Inferentia chips.
- [20]
The roughly US$7 billion figure discussed for MatX is about 75 percent above the roughly US$4 billion valuation MatX is now seeking from investors.
- [21]
Anthropic's planned Google chip purchase and its Nscale cloud deal together amount to US$81 billion in named commitments.
- [22]
The SpaceX arrangement of US$1.25 billion per month annualises to about US$15 billion a year.
- [23]
The annualised SpaceX compute payment of about US$15 billion is more than twice the roughly US$7 billion discussed for MatX.
Sources
2 independent publishers whose own reporting we read for this story.
- straitstimes.comAnthropic in talks with chip start up MatX to speed up chip design: Sources | The Straits Times
1 article · August 27, 2026
- thenextweb.comAnthropic planned a $7bn takeover of chip startup MatX, then walked away
1 article · August 28, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Custom AI acceleratorsFollow
- AI Compute ProcurementFollow
- AI mergers and acquisitionsFollow