"A horrible, demoralizing mistake." So said Joseph Bonetti, Principal Engineering Program Manager at Intel Corporation, on Linkedin earlier this week in response to rumours that TSMC might be in talks to take control of Intel's chip-manufacturing fabs.
Perhaps inevitably, Bonetti has since deleted his LinkedIn post though it is currently preserved for posterity on archive.today (via Tom's Hardware). While the post was up, Bonetti made an impassioned plea to Intel management and the Trump administration to pass up any capitulation to TSMC over chip manufacturing.
"Intel Leaders, Intel Board, Trump Administration, please do not sell out and/or give control of Intel Foundry to TSMC, just as Intel is taking a technical lead and getting out of first gear," Bonetti said.
He also laid out the case for why the rumours don't make any sense in the first place.
"Many articles say TSMC engineers will come over and share their know-how and help Intel to get their '3nm' & '2nm' nodes working. Huh? Intel's '3nm' node, Intel 3, has been in production for many months & is used in Intel's latest Xeon 6 chips.
"Intel 18A, Intel's '2nm' node is nearing completion, showing healthy progress, sampling chips to laptop makers, & is on track to be in Panther Lake chips late this year. Does TSMC have their '2nm' node, N2, up and running? No. Neither company has a '2nm' node at this point. However, Intel is on track to have theirs in production sooner," he explained.
Bonetti likewise bigged up Intel's all-important new 18A node.
"Recent reports suggest Intel 18A will outperform TSMC's N2. Both of these upcoming '2nm' nodes feature a new type of transistor called gate all around (GAA). However, InteI's 18A node has backside power delivery, an amazing engineering feat in and of itself. 18A is the more advanced node.'
"What about future nodes beyond N2 and 18A? These will likely need ASML's latest EUV tools called high numerical aperture (high NA) EUV. Intel is the only foundry that has one of these tools up and running and performing high NA R&D. Intel has two of these $350M machines up and running. It appears Intel will have at least a one-year headstart with high NA. It was reported many months ago that Intel bought up ASML's full capacity of high NA tools for 2024.
"So who would be helping who exactly if there is a joint venture?"
Without picking apart Bonetti's points in turn, part of the problem here is the difficulty comparing nodes. According to Intel's own numbers, for instance, Intel 18A has similar SRAM density to TSMC N3, not TSMC N2. In other words, some commentators, rightly or wrongly, consider Intel 18A more akin to TSMC N3 than N2, but some density measures at least.
Anyway, the real kicker in all this is the health of Intel's 18A node, on which the company's entire future has seemingly hinged for the last few years. Bonetti is obviously bullish about 18A, but perhaps surprisingly as an Intel insider, he merely says "reports suggest" 18A will outperform TSMC's N2.
Of course, we'll know soon enough if 18A is on track. Intel has committed to releasing its new Panther Lake laptop CPU on 18A later this year. So, if that's delayed or disappoints technically, the wagons really will be circling in Santa Clara.
