Chip wars reach a turning point? Reports suggest Meta (META.US) invests billions to join Alphabet Inc. Class C TPU, challenging NVIDIA Corporation's (NVDA.US) throne.

According to reports, Alphabet Inc. Class C (GOOGL.US) is significantly upgrading its competition with NVIDIA Corporation (NVDA.US) in the artificial intelligence (AI) chip race, while Meta (META.US) is emerging as a potential multi-billion dollar customer. For years, Alphabet Inc. Class C has limited the use of its custom Tensor Processing Units (TPU) to its own cloud data centers, renting them out to companies running large-scale AI workloads. However, Alphabet Inc. Class C is now marketing these chips for deployment within customers' own data centers, marking a significant shift in its strategy. One of the customers is Meta. Meta is reportedly in discussions to invest billions of dollars to integrate Alphabet Inc. Class C's TPU into its data centers starting in 2027, while also planning to rent TPUs from Alphabet Inc. Class C's cloud as early as next year. Currently, Meta primarily relies on NVIDIA Corporation's GPUs for its artificial intelligence infrastructure. With explosive growth in AI computing demand and NVIDIA Corporation's continued dominance in the supply chain, Alphabet Inc. Class C's promotion of its TPU signals a more intense phase in the AI chip war. This is significant. Executives within Alphabet Inc. Class C's cloud have stated that expanding TPU adoption could help the company capture 10% of NVIDIA Corporation's annual revenue, worth billions of dollars in profit. If this deal is reached, it would be a significant validation of Alphabet Inc. Class C's hardware ambitions. Alphabet Inc. Class C has informed potential customers - from high-frequency trading companies to large financial institutions - that installing TPUs on-site can help them meet strict security and compliance requirements for sensitive data. As a result of this news, as of the time of writing, Alphabet Inc. Class C's stock surged over 3% after hours on Monday, while NVIDIA Corporation fell over 2%. If Meta reaches an agreement with Alphabet Inc. Class C, it will help establish TPUs as an alternative to NVIDIA Corporation's chips. While NVIDIA Corporation still dominates the market, Alphabet Inc. Class C previously reached an agreement to supply as many as 1 million TPUs to Anthropic. Following the announcement of this deal, Seaport analyst Jay Goldberg stated that it was "a very strong validation" for TPU. He said, "Many people have been considering it, and now even more might be considering it." Analysts Mandeep Singh and Robert Biggar said, "Meta potentially using Alphabet Inc. Class C's TPU suggests that third-party large language model providers may consider Alphabet Inc. Class C as a secondary supplier of inference accelerator chips in the near term. We estimate that Meta's capital expenditures of at least $100 billion by 2026 mean it would spend at least $40-50 billion next year on inference chip capacity. Due to enterprise customer demand for using Alphabet Inc. Class C's TPU on the cloud and Gemini large language models, consumption and backlogs for Alphabet Inc. Class C's cloud may grow faster than other hyperscale cloud vendors and new cloud vendors." Asian stocks related to Alphabet Inc. Class C soared in early Asian trading on Tuesday. In South Korea, IsuPetasys, which supplies multi-layer circuit boards to Alphabet Inc. Class C, saw its stock price surge 18%, reaching an all-time high during trading hours. In Taiwan, MediaTek's stock rose nearly 5%. Reaching an agreement with Meta would signify a victory for Alphabet Inc. Class C. However, in the long run, this largely depends on whether TPU can demonstrate the power efficiency and computing capabilities required to be a viable choice. Alphabet Inc. Class C's TPU is a custom chip, offering advantages in cost, performance, and energy efficiency. According to reports earlier this month, Alphabet Inc. Class C's seventh-generation TPU "Ironwood" has been deployed for testing with some customers in April and will be officially available in the coming weeks. Ironwood, designed by Alphabet Inc. Class C, can handle various computing tasks from large language model training to real-time chat with Siasun Robot & Automation and AI intelligent agents. This chip can parallel up to 9,216 TPUs in a single cabinet, effectively eliminating data bottlenecks in complex models, allowing customers to run and scale the largest AI models currently available.