Anthropic is intensifying its push toward chip independence, viewing it as a critical step to solidify its competitive moat in the lead-up to a potential IPO.
According to a recent report, Anthropic previously engaged in discussions to acquire AI chip startup MatX for roughly $7 billion. However, that acquisition plan was ultimately scrapped, and the talks have since evolved into an exploration of a potential partnership. In parallel, Anthropic has also held meetings with several other AI chip startups in recent weeks, though no definitive path has been finalized yet.
These moves signal that Anthropic is seeking to reduce its reliance on external hardware providers like Nvidia, while building long-term advantages in computing costs and supply security through in-house chip design capabilities. For an AI company targeting a valuation of up to $2 trillion and preparing for a public listing, hardware autonomy has become a key part of its business narrative.
Acquisition Talks Falter, MatX Pursues Independent Funding
According to the report, citing two sources familiar with the matter, Anthropic had held in-depth negotiations with MatX over an acquisition price of approximately $7 billion. A third source indicated that the merger discussions have since shifted into talks about a collaborative relationship, though it remains unclear why the deal did not move forward.
MatX, founded by former Google TPU engineers, is now seeking to raise a new funding round at a valuation of around $4 billion. The startup's core focus lies in training chips—processors designed to build large-scale AI models. This differentiates it from some other chip startups and rival OpenAI, which have placed greater emphasis on inference chips. Sources noted that Anthropic may also pursue the development of inference chips alongside training-focused ones.
Broad Engagement with Startups and Simultaneous Talent Acquisition
Per the report, Anthropic has held talks with multiple AI chip startups over the past few weeks. These discussions are designed to help its engineers and executives gain a comprehensive understanding of the various technological approaches currently emerging in chip design, though the company has yet to make a decision on any acquisition.
On the talent front, Anthropic this week hired Amir Salek, a veteran from Google's chip division, and earlier in June brought on Clive Chan, a former OpenAI engineer who previously worked on OpenAI's chip projects.
Anthropic has stated that it is expanding its internal silicon team to design custom chips that can make its Claude models run faster and more efficiently, while maintaining a multi-supplier strategy that continues to collaborate with Nvidia and Google.
The Strategic Logic Behind Custom Chips: Lower Costs, Greater Efficiency, and Supply Risk Hedging
Anthropic's push into custom silicon is driven by several practical factors. The first is cost pressure. Although chip design is capital-intensive, with a single generation costing hundreds of millions of dollars and taking more than a year from design to usable hardware, acquiring a startup with design capabilities could significantly reduce computing costs over the long term.
The second factor is supply security. Nvidia has stated on its earnings calls that processor supply will remain tight through 2027. Anthropic has already signed multiple large-scale computing procurement agreements, including a $36 billion deal for Google AI chips, a $45 billion cloud computing lease with Nscale, and a monthly payment of $1.25 billion to SpaceX for use of its Colossus 1 data center facility, which is equipped with more than 220,000 Nvidia chips. In-house chip development could mitigate some of the uncertainty associated with external suppliers.
The third factor is performance competition. OpenAI claimed at a conference this week that its first custom chip, "Jalapeno," offers better energy efficiency for inference computing than comparable Nvidia products. Custom chips can be optimized for specific models and workloads, creating a differentiated advantage in both performance and cost-effectiveness.
Industry Trend: Major AI Labs Race to Build Proprietary Hardware
Anthropic's exploration of chip independence reflects a broader strategic direction shared by leading AI labs. Google has already developed its TPU series, while Amazon has introduced Trainium and Inferentia chips—both designed to reduce dependence on Nvidia and optimize their own AI workloads.
Anthropic was among the first AI companies to run its models across hardware from multiple vendors, giving it a foundation for a diversified hardware strategy. As its Claude family of models continues to expand, the company's demand for computing power is rising sharply.
In the context of a potential IPO, Anthropic's chip strategy carries direct significance for the capital markets. Per an earlier report, Anthropic's listing valuation target stands at $2 trillion, a figure closely tied to its projected revenue of up to $200 billion by 2028. Whether the company can establish self-reliant capabilities at the hardware level will become an important metric for investors assessing its long-term profitability potential.
Comments