Summary of Key Points
Two of the most valuable AI private companies in the world, Anthropic and OpenAI, are engaged in a battle for dominance over chip technology. Anthropic has upgraded from being a chip user to a custom chip manufacturer by securing funding from giants such as Samsung and Microsoft. OpenAI, on the other hand, has taken the lead by releasing its own custom inference chips and attempting to strengthen its position by getting the U.S. government to invest in it. Both companies are disrupting the traditional chip industry chain, forcing hardware manufacturers to adapt to their specific needs, marking a shift in AI companies’ role from being dependent on chips to defining them.
Why Did Anthropic Suddenly Decide to Develop Its Own Chips?
Anthropic’s Claude model processes a massive amount of data daily, but general-purpose GPUs (like NVIDIA’s H100) have significant drawbacks. These GPUs contain many features that are not utilized by Claude, such as training logic and graphics pipelines, making them inefficient and power-consuming. The company is also facing a severe shortage of computing resources. Anthropic’s CEO, Amoldi, stated clearly: “It’s not a lack of funds to purchase GPUs; it’s the inability to meet the growing demand for computational power.” Custom ASIC chips, designed specifically for Claude’s inference tasks, eliminate unnecessary components and optimize cost, power consumption, and speed. Previously, Anthropic had to rely on third-party solutions, but now it can design its own chips to perfectly fit its needs.
How Did Anthropic Go from a “Minor Player” to a “Power Player”?
Two years ago, Anthropic had little say in chip development. It could only suggest improvements to Amazon’s Trainium chips and was just one of Google’s largest customers, with the architecture being determined by Google. Today, things have changed:
1. Funding from Giants: In May 2026, Anthropic raised $6.5 billion, raising its valuation to $965 billion (surpassing OpenAI), with investors including Samsung, SK Hynix, and Micron—leaders in memory and chip technology. Samsung is particularly crucial as it can provide high-bandwidth memory (HBM) and 2-nanometer manufacturing services, making it an ideal partner for custom chip development.
2. Talent Acquisition: Anthropic has brought in Cliff Chen, the second-in-command of OpenAI’s hardware team who participated in the development of the Jalapeño chip, bringing expertise in both software and hardware.
3. Diversified Partnerships: The company is not limited to Samsung; it is also in talks with Microsoft for Maia chips and Fractile in the UK for inference chips, aiming to reduce reliance on a single supplier and establish its own chip standards.
OpenAI’s Counterattack: Chips + Government Support
OpenAI has moved faster than Anthropic:
1. Release of Jalapeño Chips: These custom chips were developed specifically for large-scale model inference and were completed in just nine months (traditional high-end chips take 1-2 years). They are designed solely for inference tasks, significantly reducing costs. Broadcom’s CEO stated that Jalapeño will be piloted with GPT-5.3 by the end of 2026 and scaled up in 2027.
2. Government Investment: OpenAI is seeking a 5% stake from the U.S. government (worth approximately $42.6 billion) and has proposed similar arrangements with Google and Anthropic, creating a public fund where AI company profits would be shared with the public. This move not only gains public support but also gives OpenAI governmental backing, putting it at an advantage over Anthropic (considering that GPT-5.6’s release was delayed due to government requests, and Anthropic’s Fable 5 was once banned).
The Essence of this Competition: AI Companies Becoming Rule-Makers
The traditional chip industry chain followed the pattern of “chip manufacturers design → manufacturing → AI companies purchase.” However, with AI companies spending billions on computing power annually, they are now taking control of the design process. For example, Anthropic is asking Amazon to build data centers and Google to supply TPU chips, while OpenAI is influencing NVIDIA, AMD, and Cerebras to meet its needs. This shift means that:
- AI companies have more influence: Hardware manufacturers must adapt their technology to AI companies’ requirements (e.g., Samsung optimizing 2-nanometer processes for Anthropic).
- Reconfiguration of the Industry Chain: Traditional chip giants like NVIDIA are no longer the sole players; AI companies are becoming the new leaders.
- Intense Competition Ahead: The company that dominates chip technology will have a lower cost and faster model iteration, giving it a competitive edge in the AI race.
This battle is just beginning. Once Anthropic and OpenAI go public, the global tech landscape could be completely transformed.
(The entire analysis is explained in plain language to make the logic of these AI giants’ chip wars understandable to everyone.)