Intel wins Google validation
Google has committed to using multiple generations of Intel AI chips in its data centres, a public signal of confidence that bolsters Intel’s comeback narrative. Intel is also preparing EMIB‑T packaging for rollout and has seen strong investor enthusiasm, but its foundry arm still shows steep losses — a reminder that execution and cost risk remain (cnbc.com) (tomshardware.com) (businesstoday.in).
Google just gave Intel something it has been missing in the artificial intelligence race: a public customer saying it will keep buying Intel chips for future data centers, not just the current generation. Google and Intel said on April 9 that they are expanding a multiyear partnership across multiple generations of Intel Xeon processors. (cnbc.com) (intel.com) That matters because the artificial intelligence boom made graphics chips the stars, but data centers still need central processors to move data, schedule work, and keep whole systems running. Intel and Google said central processors will continue powering Google Cloud infrastructure for artificial intelligence, inference, and general-purpose workloads. (intel.com) (reuters.com) Google is not starting from zero here. CNBC said Google has relied on Intel processors since its earliest server-rack buildout nearly three decades ago, and the new deal extends that relationship into the part of the market where Intel most needs proof it still belongs. (cnbc.com) The partnership is also wider than ordinary chip supply. Intel said the two companies will co-develop custom infrastructure processing units, which are specialized chips that handle the traffic-cop jobs inside giant computing networks, so expensive artificial intelligence hardware is used more efficiently. (intel.com) (reuters.com) Intel’s second bet sits one layer lower, in packaging, which is the method for wiring several tiny chips together so they act like one bigger product. Tom’s Hardware reported that Intel plans to roll out its Embedded Multi-die Interconnect Bridge-T, or EMIB-T, packaging into fab production this year, aiming at systems that need high-bandwidth memory and chiplet-style designs. (tomshardware.com) That packaging matters because the fastest artificial intelligence systems are no longer one giant slab of silicon. They are more like a city block of smaller pieces, with compute tiles and memory stacks linked by ultra-short bridges, and Intel is pitching EMIB-T as the plumbing for that layout. (tomshardware.com) (trendforce.com) Investors heard the Google announcement as evidence that Intel’s comeback story is not just a slide deck. Reuters said the deal comes as companies shift from training giant models toward deploying them, a phase that increases demand for central processors, and market coverage on April 9 showed Intel shares rising after the news. (reuters.com) (fool.com) But the hard part of Intel’s turnaround is still its factory business. Intel’s full-year 2025 results showed Intel Foundry lost $10.3 billion on $17.8 billion in revenue, and outside customers contributed only $307 million, which means most of the business is still supported by Intel’s own internal chip demand. (intel.com) (fool.com) So this is the picture now: Google has endorsed Intel’s central processors for future artificial intelligence data centers, Intel is trying to sell the shovels and the pipes through advanced packaging, and the company still has to prove it can turn those wins into profitable factory volume. The comeback story is no longer missing customers, but it is still missing the easy part where the math works. (cnbc.com) (tomshardware.com) (intel.com)
