I suspect Pat Gelsinger has thought about what Intel should (and shouldn’t) do since he left in ‘09. He’s had the benefit of being inside for 30 years, and then 12 years on the outside watching the company. No one could have a better perspective and understanding from the transistor level to the macroeconomics of the 80 billion global company.
Murthy Renduchintala gets credit for breaking down the silo walls of the x86 cathedral and letting in the light of the GPU in 2017. That was a start, an important start. Next came a heterogeneous processor API Intel calls OneAPI and along with it enhanced and expanded packaging techniques and technology under the umbrella heading of Foveros.
Intel led the U.S. State-of-the-Art Heterogeneous Integration Prototype (SHIP) program, and in 2020 won the award to develop prototypes of multichip packages and accelerate the advancement of interface standards, protocols, and security for heterogeneous systems. SHIP prototypes will integrate special-purpose government chips with Intel’s advanced, commercially available silicon products, including field-programmable gate arrays, application-specific integrated circuits, and CPUs. And most recently Intel announced plans to buy Israeli-based Tower Semiconductor to “help scale Intel’s foundry services and advance our goal of becoming a major provider of foundry capacity globally,” said Gelsinger.
And then Gelsinger announced Intel would, as part of his Integrated Device Manufacturer (IDM) 2.0 strategy, build chips with RISC-V, or Arm, or FPGA, a GPU, tensor-cores, x86, or just about any processor that has an RTL library. Not only that but Intel will put the heterogeneous processors in a single package for customers and wrap it up with a nice OneAPI bow.
|Intel builds, mounts, and delivers any processor desired. (Source: Intel)|
Furthermore, not only can Intel do that at scale, but it can also do that at nodes, the tiniest nodes there are.
This is an audacious big idea that will rock Intel to its very foundations, delight its shareholders, thrill all the engineers and technicians in Intel, delight its customers, and scare the bejeezus out of its competitors. No other company other than maybe IBM or Toshiba could take on such a grand plan. It takes deep deep pockets, a very long horizon, a thorough understanding of the many many moving parts in launching such a plan, and the courage to see it through.
This is the new Intel.
What do we think?
Intel’s BOD deserves a pat or two on the back for (a) (finally) waking up, and (b) having the courage to back a super-smart rebel who clearly thinks outside the box that Intel had trapped itself in. The company has stacked up IP over the past two decades, owns every imaginable AI processor, has vision systems, communications, memory, and now GPUs—real GPUs, not the memory and power-starved iGPUs it has been building (which perform remarkably well given the constraints they are placed under).
The company is at the forefront of the U.S.’s work about the importance of semiconductors—it only took a trade war and a pandemic to get its attention on that.
But it will take time to get all these super complex and complicated pieces running at speed and in synch. That’s where the courage comes in. I imagine the phrase Gelsinger uses the most in the meeting he attends is: “Wait for it,… wait for it….”
One of the signals that Wall Street gets it will be when Hathaway buys in. Wait for it…