While Semiconductor Stocks Plunge, Anthropic Plans to Develop a 2nm Chip

链捕手Publicado a 2026-07-03Actualizado a 2026-07-03

Resumen

Anthropic, the AI company behind Claude, is exploring the development of its own custom AI chip, according to a report from The Information. The company is in early discussions with Samsung Electronics to manufacture the chip using Samsung's most advanced 2-nanometer process and packaging technology. While the project is still in preliminary stages, including defining chip specifications, and could be abandoned, it marks a strategic step for Anthropic. The move comes as the company seeks greater control over its computing costs and hardware optimization, particularly for inference tasks to run its models more efficiently and cheaply. Samsung's potential involvement follows its participation as a strategic investor in Anthropic's recent $65 billion funding round. For Samsung, partnering with a major AI lab represents a significant opportunity for its foundry business to compete with market leader TSMC in advanced semiconductor manufacturing. Anthropic's CEO, Dario Amodei, has previously highlighted the immense financial challenge of securing enough computing power for anticipated growth, making cost-effective inference a critical focus. The company would join other tech giants like Google, Amazon, Microsoft, Meta, and OpenAI in pursuing custom AI silicon. However, analysts note this trend creates deeper interdependencies rather than independence, as US AI labs become more tightly woven into Asian semiconductor supply chains. Despite this move, Anthropic remains heavily reli...

By Bo Yang, Tencent Tech

Anthropic is building its own chip, and like Musk, has chosen Samsung over TSMC.

The Information, citing three informed sources, reported that Anthropic has initiated early-stage work on a self-developed AI chip and is in discussions with Samsung to manufacture this chip using Samsung's most advanced 2-nanometer process and packaging technology.

The aforementioned sources said that Anthropic is also defining the chip's functions, performance metrics, and its integration scheme within servers or clusters. Currently, the company has begun discussions with several design service providers, but it is far from the detailed design, testing, and mass production stages. The sources also cautioned that Anthropic could still abandon the project midway.

Public information shows that Anthropic's chip team has begun forming. For instance, on June 9th, the company poached Clive Chan, an early member of OpenAI's custom chip team.

Clive Chan, an early member of OpenAI's custom chip team, has joined Anthropic

Anthropic has been extremely cautious in its external statements regarding self-developed chips, only emphasizing that future computing power expansion will still primarily rely on AWS's Trainium, Google's TPU, and NVIDIA's GPUs, while declining to comment on its own chip roadmap.

In fact, rumors about Anthropic's plans to develop its own chip emerged months ago. Reuters reported in April that Claude's computational demands were growing too rapidly, outpacing available computing power, leading Anthropic to consider building its own chips. However, at that time, it was just a preliminary consideration. Now, it appears the matter has moved a step forward.

It should be noted that the disclosure of Anthropic's entry into chip-making coincides with a two-day plunge in global semiconductor stocks. The Philadelphia Semiconductor Index fell nearly 12% over two sessions, its largest drop since June 5th. Memory chip stocks were hit particularly hard, with Samsung Electronics falling 9.1% and SanDisk dropping 14.1% overnight.

01 A Foreshadowing from a Funding Round, and Samsung Foundry's Breakout Battle

Samsung becoming Anthropic's partner is not surprising.

In May of this year, Anthropic completed a $650 billion Series H funding round, with Samsung Electronics participating as a strategic investor, alongside other memory chip giants SK Hynix and Micron Technology.

At that time, memory chip supply was tight, forcing consumer electronics manufacturers like Apple to raise prices. A direct benefit of this investment was binding Anthropic to the chip suppliers critical to its growth.

But Samsung's calculations extend beyond financial returns.

The company is a king in the memory chip sector, but has consistently lagged behind TSMC in the foundry business. TSMC's advanced production lines remain the industry standard for cutting-edge AI chips, but the current demand for AI chips is straining TSMC's capacity. This provides Samsung an opportunity to win customers with its 2nm technology.

According to The Information, Google is already considering having Samsung manufacture some of its future TPUs. If Samsung can also secure Anthropic, it would be a significant victory for its foundry business.

Two informed sources revealed that Anthropic is evaluating Samsung's 2nm manufacturing process and advanced packaging facilities. "2nm" is one of the most advanced chip manufacturing technologies, making processors denser and more energy-efficient. Advanced packaging places the main processor closer to high-speed memory, allowing data to travel faster within the chip.

Samsung is NVIDIA's manufacturing partner for HBM memory and LPU chips; conversely, Samsung relies on NVIDIA's software for chip production. The two companies are jointly building an AI chip factory in South Korea. If Anthropic joins the customer list, Samsung gains another chip in its efforts to catch up to TSMC in advanced process nodes.

Samsung's confidence in courting Anthropic is backed by South Korea's national-level bet on the semiconductor industry.

The South Korean government recently announced a decade-long investment plan led by the Samsung Group (Samsung Electronics' parent company) and SK Group (SK Hynix's parent company), with a combined investment of $518 billion to build four memory chip factories in South Korea.

Samsung alone announced on July 2nd an investment plan for the Chungcheong region exceeding 140 trillion won, approximately $90 billion. Under this plan, Samsung Electronics will build next-generation HBM production bases in Cheongju and Wonyang, adding five production lines. Since the AI boom, HBM has been in short supply; expansion could increase Samsung's revenue by tens of billions of dollars in the coming years. Samsung Electro-Mechanics is also expanding AI server packaging substrate capacity in Sejong City while simultaneously investing in advanced packaging technology R&D.

By participating in Anthropic's funding round, Samsung is betting on becoming the latter's long-term hardware partner.

02 Inference Chips: A Breakthrough Point for a Costly Cycle

Anthropic's first 2nm chip, like OpenAI's, is expected to focus on inference.

Custom inference chips would allow Anthropic's models to run faster, consume less power, reduce cloud bills, and be specifically hardware-optimized for Claude.

Previously, OpenAI selected Broadcom to design an inference chip codenamed "Jalapeño," aiming to run large language models more efficiently. OpenAI claimed the chip is more energy-efficient with better performance per watt than competitors.

OpenAI partners with Broadcom to design an inference chip codenamed "Jalapeño"

Google has TPU, Amazon has Trainium, Microsoft has Maia, and Meta is also working on its own. Anthropic is almost the last major AI lab still fully reliant on others' chips. The allure of autonomy in this area is too strong.

In February of this year, Anthropic CEO Dario Amodei, in an interview with podcast host Dwarkesh Patel, discussed the logic behind compute strategy.

Amodei said: "If I assume revenue continues to grow 10x per year, then by the end of 2026 it's $100 billion, and by the end of 2027 it's $1 trillion. I could order $1 trillion worth of compute based on that. But what if I'm wrong? What if the growth multiplier is 5x instead of 10x? If revenue in 2027 is only $800 billion instead of $1 trillion, then I'm bankrupt. No hedge can save me."

Therefore, Amodei must find a balance for Anthropic: purchase enough compute to handle high-growth scenarios, but avoid over-expansion to the point where a slowdown in growth would crush the company under its upfront investments.

Amodei revealed that the global compute capacity built this year is roughly 10 to 15 gigawatts, approximately tripling annually. By 2028 or 2029, the industry's annual compute investment will reach several trillion dollars. He believes this is the trend the industry is following.

Developing its own chip is like raising the stakes in this gamble.

On one hand, any efficiency gain in inference from a self-developed chip can save substantial costs across clusters at the gigawatt scale. On the other hand, when everyone is scrambling for processors, data centers, and power, having its own chip provides extra leverage in negotiations.

03 Supply Chain "Nesting Doll": Who Depends on Whom?

Anthropic desires chip independence, but this web is becoming increasingly intricate.

Programmer Benny Lam pointed out a phenomenon on social media: every US AI lab is collaborating with Asian foundries for custom chips. The result is not independence but tighter entanglement of the supply chain.

Anthropic needs Samsung's fabs to reduce reliance on NVIDIA. NVIDIA needs Samsung's fabs to meet its expanding capacity demands. And Samsung needs to retain both major customers simultaneously to keep its foundry business competitive against TSMC.

The three are bound together; none can leave the others.

Lam commented that the trend towards custom chips was supposed to give AI labs more independence, but each new collaboration between a US lab and an Asian foundry makes this network denser and harder to disentangle.

Tech blogger Vaibhav Sisinty also noted Samsung's unique role in this chess game.

He wrote that Samsung's investment in Anthropic's $650 billion funding round wasn't just for financial returns. Samsung is a chip manufacturer, memory supplier, advanced packaging factory. Its investment bets on becoming Anthropic's hardware partner. With millions of queries daily, Anthropic pays for compute resources each time—its bill is astronomical. Whoever controls its own inference chip controls its scaling speed and cost.

04 NVIDIA's Fortress Remains Formidable

Designing a competitive AI chip is extremely complex. Architecture, software, manufacturing, testing, large-scale deployment—each step requires significant time. Anthropic cannot abandon its current hardware partners in the short term.

Looking at Anthropic's current compute layout, its strategy is clear. The company has consistently adhered to a multi-cloud leasing approach, unlike OpenAI or xAI which have individually tied themselves to NVIDIA hardware. It uses chips from Amazon, Google, and NVIDIA, and is negotiating chip cooperation with Microsoft and the UK startup Fractile.

Anthropic's ties with Amazon and Google are particularly deep.

In April 2026, Anthropic expanded its collaboration with AWS, committing to invest over $100 billion in the next decade, securing 5 gigawatts of new compute capacity, including custom chips like Trainium2, Trainium3, and Trainium4. Amazon's cumulative investment in Anthropic has reached tens of billions of dollars.

Simultaneously, Anthropic signed long-term agreements with Google and Broadcom to secure approximately 3.5 gigawatts of next-generation TPU compute starting in 2027. Previously in 2025, the two parties had already signed agreements worth hundreds of billions of dollars, involving millions of TPUs and over 1 gigawatt of capacity.

Claude is also one of the few frontier models simultaneously available on AWS, Google Cloud, and Azure.

And that's not all. In May 2026, Anthropic leased the entire capacity of the SpaceX Colossus 1 compute cluster from xAI.

Located in a Memphis data center, this cluster houses over 220,000 NVIDIA GPUs, consuming over 300 megawatts of power, with a monthly rent of approximately $1.25 billion. The total contract value is hundreds of billions of dollars, running until around 2029. The two parties have even discussed deploying multi-gigawatt compute capacity in orbit.

Wide-ranging deployment and multi-vendor parallelism is Anthropic's way of spreading risk.

But for now, NVIDIA remains the house in this game. According to The Information's estimates, despite the ongoing financing and design activities in the inference chip market, NVIDIA's market share has actually risen to 74% in recent years. CEO Jensen Huang consistently asserts that NVIDIA's chips are more efficient than any alternative for inference tasks.

Criptos en tendencia

Preguntas relacionadas

QAccording to the article, why is Anthropic considering developing its own AI chip, and who is its potential manufacturing partner?

AAnthropic is considering developing its own AI chip to improve the speed and power efficiency of running its models (like Claude) and to reduce cloud costs. Its potential manufacturing partner is Samsung, specifically for using Samsung's most advanced 2-nanometer process and packaging technology.

QWhat strategic goal is Samsung aiming to achieve by partnering with Anthropic on chip development and having invested in the company?

ASamsung's strategic goal is to win a significant customer for its chip foundry business, helping it compete with the industry leader TSMC. By partnering with Anthropic, Samsung gains a foothold in the advanced AI chip market and strengthens its position in the global semiconductor supply chain.

QWhat type of chip is Anthropic's first 2nm project expected to focus on, and what are the potential benefits?

AAnthropic's first 2nm chip project is expected to focus on inference chips. The potential benefits include running large language models faster, with lower power consumption, and at a reduced cost, while also allowing for hardware specifically optimized for its Claude model.

QHow does the article describe the current interdependency in the AI chip supply chain, particularly involving Anthropic, Samsung, and Nvidia?

AThe article describes a complex interdependency: Anthropic needs Samsung's foundry to reduce reliance on Nvidia, Nvidia needs Samsung's foundry to meet its own capacity demands, and Samsung needs to retain both Anthropic and Nvidia as major customers to keep its foundry business competitive against TSMC. This creates a tightly knit network where each party relies on the others.

QDespite exploring custom chips, what is Anthropic's current strategy for securing computing power, and which companies are its key providers?

AAnthropic's current strategy is a multi-cloud, multi-vendor approach to secure computing power and分散风险. Its key providers are Amazon Web Services (using Trainium chips), Google Cloud (using TPUs), and Nvidia (GPUs). It also has significant contracts with these companies and has even leased a large GPU cluster from xAI.

Lecturas Relacionadas

Hot Takes|Why Did the Famous "Tech Lead" Dump All His Bitcoin? The "Investment Whiz Kid" is Here!

**Weekly Spicy Review: Tech Lead's Bitcoin Bust, Reddit Meme, and Trump's Crypto Cash** This week's "Spicy Review" covers three notable incidents from the crypto world. **1. A Tech Lead Learns the Hard Way:** A former Google and Meta technical lead, Patrick Shyu, went viral after revealing he was forced to liquidate all his Bitcoin holdings. He suffered massive losses due to excessive leverage during Bitcoin's sharp decline from $120k to $60k. He shared critical observations: crypto trading often hinges on attention, not fundamentals; Bitcoin lacks a stable source of public focus; the AI boom is diverting capital; and Bitcoin faces structural risks like centralization of code maintenance and quantum computing threats. Despite his short-term exit, he remains a long-term believer. **2. Reddit Roasts the "Investment Whiz":** A popular meme on Reddit's CryptoCurrency subreddit depicted MicroStrategy's Michael Saylor looking down from a balcony. The caption joked about his relentless focus on buying Bitcoin with corporate funds, contrasting with average investors' mundane concerns. The post sparked humorous commentary on his high-risk, high-conviction strategy. **3. Trump's $1.4 Billion Crypto Haul:** The White House's financial disclosure revealed former President Donald Trump earned at least $1.4 billion from cryptocurrency activities in a year, contributing to a total income of over $2.2 billion. This windfall stands in stark contrast to the performance of "TrumpCoin" (officially DJT), which plummeted over 97% from its peak, reportedly causing investor losses exceeding $2 billion. Critics, like California Governor Gavin Newsom, accused Trump of profiting while his supporters suffered losses. The week highlighted a mix of painful lessons learned from leverage, community humor at industry figures, and the stark realities of political figures capitalizing on the crypto market.

Foresight NewsHace 27 min(s)

Hot Takes|Why Did the Famous "Tech Lead" Dump All His Bitcoin? The "Investment Whiz Kid" is Here!

Foresight NewsHace 27 min(s)

Trading

Spot

Artículos destacados

Cómo comprar CHIP

¡Bienvenido a HTX.com! Hemos hecho que comprar USD.AI (CHIP) sea simple y conveniente. Sigue nuestra guía paso a paso para iniciar tu viaje de criptos.Paso 1: crea tu cuenta HTXUtiliza tu correo electrónico o número de teléfono para registrarte y obtener una cuenta gratuita en HTX. Experimenta un proceso de registro sin complicaciones y desbloquea todas las funciones.Obtener mi cuentaPaso 2: ve a Comprar cripto y elige tu método de pagoTarjeta de crédito/débito: usa tu Visa o Mastercard para comprar USD.AI (CHIP) al instante.Saldo: utiliza fondos del saldo de tu cuenta HTX para tradear sin problemas.Terceros: hemos agregado métodos de pago populares como Google Pay y Apple Pay para mejorar la comodidad.P2P: tradear directamente con otros usuarios en HTX.Over-the-Counter (OTC): ofrecemos servicios personalizados y tipos de cambio competitivos para los traders.Paso 3: guarda tu USD.AI (CHIP)Después de comprar tu USD.AI (CHIP), guárdalo en tu cuenta HTX. Alternativamente, puedes enviarlo a otro lugar mediante transferencia blockchain o utilizarlo para tradear otras criptomonedas.Paso 4: tradear USD.AI (CHIP)Tradear fácilmente con USD.AI (CHIP) en HTX's mercado spot. Simplemente accede a tu cuenta, selecciona tu par de trading, ejecuta tus trades y monitorea en tiempo real. Ofrecemos una experiencia fácil de usar tanto para principiantes como para traders experimentados.

428 Vistas totalesPublicado en 2026.04.21Actualizado en 2026.06.02

Cómo comprar CHIP

Discusiones

Bienvenido a la comunidad de HTX. Aquí puedes mantenerte informado sobre los últimos desarrollos de la plataforma y acceder a análisis profesionales del mercado. A continuación se presentan las opiniones de los usuarios sobre el precio de CHIP (CHIP).

活动图片