TSMC To Invest Up To $1B in Anthropic And OpenAI – Semiecosystem

Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
By Mark LaPedus
TSMC’s board has approved a measure under which the company will invest up to $1 billion in two major AI companies–Anthropic and OpenAI.
In one move, TSMC, the world’s largest foundry vendor, will make an investment in Anthropic in an amount not to exceed $500 million. Then, in another move, TSMC will invest in OpenAI in an amount not to exceed $500 million.
The investment in Anthropic is particularly noteworthy. Anthropic, a San Francisco-based AI company, last week launched its proposed initial public offering (IPO). Anthropic has applied to list its Class A common stock on Nasdaq under the symbol “ANT.” With the IPO, Anthropic is expected to have a valuation of $1.4 trillion.
As part of the IPO, AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC have separately agreed to purchase an aggregate of $26 billion in Anthropic’s common stock in a private placement, according to a filing from Anthropic in the U.S. Securities and Exchange Commission (SEC).
For its part, TSMC will acquire Anthropic’s common shares at its per share IPO price in an amount not to exceed $500 million. Then, in a separate move, TSMC will acquire OpenAI’s preferred stock in an amount not to exceed $500 million.
Anthropic and OpenAI are separately developing complex and powerful generative AI models, which can be used for analysis, coding and R&D. These AI models are also designed to solve the most daunting technical challenges.
On a PC, smartphone or other system, consumers, businesses and researchers can access so-called AI chatbots, or conversational AI assistants, from Anthropic and OpenAI. OpenAI’s AI chatbot is called ChatGPT, while Anthropic’s technology is known as Claude. In general, AI chatbots or assistants can generate text, images, videos, audio, computer code or other forms of digital data.
TSMC is directly or indirectly involved with both Anthropic and OpenAI. Basically, TSMC manufacturers various AI chips from different vendors. In a system, these chips process these complex AI models from Anthropic, OpenAI and other AI companies.
In addition, Anthropic and OpenAI are also developing their own AI chips, which may–or may not–be manufactured by TSMC in the future. TSMC is investing in both companies to help prop up two potential AI foundry customers. And in the distant future, the foundry giant would like to get a return on investment.
To be sure, Anthropic and OpenAI are important companies in the AI world. So, let’s take a look at both companies, and find out where AI technology is heading.
The great AI debate
But before describing Anthropic and OpenAI, let’s look at AI and briefly talk about its pros and cons. Needless to say, AI is a hot topic. The AI proponents insist the technology will benefit mankind. AI can enable new breakthroughs in science, medicine and other fields. It can boost productivity for a growing number of tasks.
AI has also taken some hits. First, some have warned about the dangers of AI. Then, the technology is already making some job types obsolete. And there is a growing backlash about the development of AI data centers in many communities.
Still, AI companies are racing each other to build more data centers. The combined capital spending among the world’s nine leading cloud service providers is projected to surge 106% year-over-year to $957 billion in 2026,” according to TrendForce (See Figure 1 below).
Many wonder if this spending is out of control. Others believe the AI market will soon slow, if not hit the wall.
Then, on the geopolitical front, the U.S. and China are in a race to develop the most powerful AI models and chips. It’s unclear where the AI race is heading. And if that isn’t enough, Jensen Huang recently resigned as president and chief executive of Nvidia, the world’s largest AI chip supplier with 90% market share. Huang, the visionary of Nvidia, is also the face of AI. This in turn creates more uncertainty in the mix.
Figure 1. AI CapEx forecast for the nine leading CSPs. The world’s nine leading CSPs are Google, Amazon, Meta, Microsoft, Oracle, ByteDance, Tencent, Alibaba and Baidu. Source: TrendForce
Who is OpenAI?
Nonetheless, we can’t ignore AI. We need to get a better understanding of some of the key AI players and where AI technology is heading.
So let’s take a look at both Anthropic and OpenAI. Based in San Francisco, OpenAI was founded in 2015. At the time, Elon Musk and Sam Altman were co-chairs. Musk resigned in 2018. At present, Altman is OpenAI’s chief executive.
OpenAI changed the world in 2022, when it launched ChatGPT. Suddenly, ChatGPT became a huge hit among consumers, which in turn accelerated the worldwide AI boom. In May 2026, ChatGPT reached 1 billion monthly active users, according to Sensor Tower, a provider of digital intelligence.
Last month, OpenAI introduced its latest AI model. This technology, called GPT-6.1 Terra, is a faster AI model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing.
The technology behind ChatGPT and competitive AI models is called a large language model (LLM). Developed by AI companies, “a large language model (LLM) is a model designed to work with language,” according to OpenAI. “It learns patterns from large amounts of text from many sources so it can generate and transform text in helpful ways. An LLM doesn’t ‘know’ things the way a person does. Instead, it predicts the most likely next piece of language based on context. Over time, advances in computing power, training methods, and access to large datasets made it possible to build larger and more capable large language models.”
Generally, to develop an LLM, an AI company will train the model in two stages. The first step, called the pre-training stage, is when the model learns general patterns from a huge amount of data, according to OpenAI. Then, the second step, called the post-training stage, “helps the model follow instructions more reliably, communicate in a useful style, and handle tricky situations better,” they added.
OpenAI is developing among the world’s most powerful LLMs. But at present, the company doesn’t own or operate any data centers. Instead, the company leases computing space from various AI data center operators, sometimes called cloud service providers (CSPs). Amazon, Google, Microsoft and others are CSPs.
In general, CSPs own and operate large data centers. Each facility consists of a multitude of high-speed computers, usually called servers. These servers incorporate AI chips and other devices, which are designed to process and accelerate the latest AI models.
Meanwhile, in recent times, OpenAI has revised its strategy. OpenAI, along with various partners, is in the process of building several of its own data centers. For example, OpenAI is building a giant data center in Effingham County, Georgia. To support the data center, the company is contracting with Georgia Power Company for 3.2 gigawatts of power. To put this into perspective, a 1-gigawatt facility “consumes the power equivalent of a major U.S. city or a full-scale nuclear reactor,” according to the Global Data Center Hub, a technology site.
For its data center projects, OpenAI is working with several AI chip suppliers, including AMD, Broadcom, Cerebras and Nvidia.
OpenAI is also developing its own AI accelerator chip, dubbed Jalapeño. OpenAI will continue to use third-party AI chips. But Jalapeño is optimized for OpenAI’s own LLMs. Thus, the company may use the Jalapeño AI chip in select LLM applications. And with Jalapeño, OpenAI isn’t totally dependent on AI chips from others.
The company is co-developing Jalapeño with Broadcom (stock symbol: AVGO). Broadcom is also coordinating the manufacturing process through TSMC. “Expectations are for OpenAI’s Jalapeño AI ASIC, which has been won by AVGO, to begin ramping in 2H27 into 2028 with next year’s volume expected to be 300K and total lifetime volumes of 1M units,” said John Vinh, an analyst with KeyBanc Capital Markets, in a recent research note.
To be sure, OpenAI is a fascinating company to watch, but it faces several challenges. OpenAI generates revenue through a multi-tiered subscription model. But the company is unprofitable and continues to lose huge sums of money. Then, it recently delayed its IPO. And its data center project in Georgia has sparked a community backlash.
Who is Anthropic?
Meanwhile, in 2021, Anthropic was founded by former members of OpenAI. Then, in 2023, Anthropic released Claude, a next-generation AI assistant. When it was introduced, Claude was designed for several applications, including summarization, search, collaborative writing, Q&A and coding.
Anthropic doesn’t own or operate any data centers. Instead, the company leases computing space from various CSPs, such as Amazon, Google, Microsoft and others, according to Anthropic’s IPO filing.
Within its leased space at the various CSPs, Anthropic makes use of a multitude of servers to process its AI models. Depending on the CSP, these servers incorporate various AI chip architectures, such as Amazon’s Trainium AI chips, Google’s Tensor Processing Units (TPUs), Nvidia’s GPUs and AMD’s GPUs, according to the filing. For the most part, these AI chips are manufactured by TSMC.
Anthropic is also working on its own AI chip. To design the AI chip, the company is looking for a co-development partner. Two companies, Broadcom and MediaTek, are competing for Anthropic’s business as co-development partners. Intel Foundry is also competing for the foundry business. TSMC also wants the foundry business.
“We learned that Anthropic also is working on an AI ASIC with the finalists for the ASIC provider being down to AVGO and MediaTek. The program scale is significant with lifetime volumes expected to be 5M-10M units and ramp at the end of 2028/early-2029,” KeyBanc’s Vinh said. “On the foundry side, INTC is attempting to win this on 18AP, as well with EMIB-T.”
Meanwhile, last week, Anthropic had two major announcements. First, Anthropic launched its IPO. The proposed offering consists of 825,000,000 shares of Class A common stock. The IPO price is expected to be between $70.00 and $78.00 per share.
Anthropic generates revenue through enterprise contracts and paid subscriptions. According to company’s filing, Anthropic’s sales were $4.585 billion in 2025, up 1,088% over 2024. It also reported a net loss of $41.965 billion in 2025. For the first nine months of 2026, the company reported a loss of $104.039 billion on sales of $33.242 billion.
Last week, Anthropic also introduced two new versions of its AI models–Claude Fable 5.2 and Claude Mythos 5.2. For most coding and knowledge tasks, Anthropic recommends that users start with Claude. Claude Mythos 5.2 is designed to find vulnerabilities in cybersecurity applications.
Both AI LLMs are up to 30% faster than previous versions. During prerelease testing, Claude Fable 5.2 and Claude Mythos 5.2 enabled the company’s engineers to complete infeasible software migrations, designed three working electronic devices from scratch and solved two open problems in the physics of black holes. The LLMs formalized the proof of Fermat’s Last Theorem, a complex mathematical formula, in 30 hours.
Dario Amodei, chief executive of Anthropic, believes that AI could cure most major diseases in the next 5 to 10 years. AI has other benefits, as well.
Amodei has also been outspoken about the dangers of AI, especially without guardrails or regulation. “But like many technologies before it, AI brings risks, and because it is such a powerful technology, these risks are serious. I’ve written a lot about them too. They include the risk of losing control of AI systems, misuse of AI for cyberattacks and bioterrorism, and serious economic disruption. A race to the bottom, spurred by commercial incentives, can make these risks more acute,” Amodei wrote in a recent blog.
“(Over) the last few months, I have become convinced that fully addressing the risks requires even more prudence — not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up. We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain,” he added.
It, however, doesn’t appear Anthropic or other AI concerns are slowing down. In fact, the technology is speeding up, thereby keeping us on edge. The AI community wants us to cross our figures and hope for the best.

source

Scroll to Top