AI compute market broadens beyond GPUs
TSMC reported record preliminary Q1 revenue on continued AI-chip demand even as industry attention widens beyond accelerators to the 'plumbing' of AI infrastructure. At the same time Intel and Google expanded a multi‑year partnership to co-develop data‑centre infrastructure and custom chips — signalling renewed emphasis on CPUs, infrastructure processing units, and system-level efficiency — while Nvidia’s shares continued to climb on hyperscaler capex optimism. (Simply Wall St, Indian Startup News, (tradingview.com))
Artificial intelligence chip demand is still lifting the sector, but the money is spreading beyond graphics processors into the servers, networking and control chips that keep data centers running. (cnbc.com) Taiwan Semiconductor Manufacturing Co. said April 10 that first-quarter revenue rose 35% from a year earlier to NT$1.13 trillion, or about $35.6 billion, above LSEG estimates of NT$1.12 trillion. March revenue alone climbed 45.2% to NT$415.2 billion. (cnbc.com) The foundry, which builds chips for Nvidia and Apple, had guided first-quarter revenue to $34.6 billion to $35.8 billion and is scheduled to report full quarterly earnings on April 16, 2026. Its investor site shows first-quarter actual revenue landed near the top end of that range. (investor.tsmc.com) A data center does not run on accelerators alone. Central processing units coordinate work, and infrastructure processing units act like traffic cops for networking, storage and security so the expensive graphics processors spend more time doing artificial intelligence jobs. (newsroom.intel.com) Intel and Google said April 9 they are expanding a multiyear partnership around Intel Xeon central processing units and custom infrastructure processing units for Google’s cloud infrastructure. Intel said the chips will target artificial intelligence, inference and general-purpose workloads across Google’s fleet. (newsroom.intel.com) Google Cloud already uses Intel Xeon 6 processors in its C4 and N4 instances, and the companies said the new work will span multiple processor generations. Intel said the goal is lower total cost of ownership and better energy efficiency as artificial intelligence systems become more complex. (newsroom.intel.com) Nvidia is still the market’s clearest proxy for that spending boom. The company said in February that data center revenue jumped 75% from a year earlier to $62.3 billion, while networking revenue rose 263% to $10.98 billion. (cnbc.com) That networking figure is a reminder that even Nvidia’s growth is not just about graphics processors. Its NVLink and Spectrum-X products connect and move data between hundreds of chips inside artificial intelligence clusters. (cnbc.com) Wall Street has kept rewarding that buildout. Through April 11, Nvidia shares had risen for eight straight sessions, their longest winning streak since November 2023, as investors focused on chip demand and cloud capital spending. (finance.yahoo.com) The result is a broader artificial intelligence hardware race: Taiwan Semiconductor Manufacturing Co. is filling more advanced orders, Intel is pushing central and infrastructure chips back into the conversation, and Nvidia is selling both processors and the links between them. The next test comes with Taiwan Semiconductor Manufacturing Co.’s full earnings report on April 16. (investor.tsmc.com)
