Analysis With its $5.4 cardinal bid to get Tower Semiconductor successful ruins, Intel Foundry Services' (IFS) maestro scheme has been turned connected its head.
The failed buyout would not only person unlocked a bounty of mature process tech to compliment Intel's leading-edge nodes, but besides much-needed expertise successful really to tally a successful foundry business.
While Intel has operated its ain fabs for decades, nan silicon slinger doesn't person immoderate acquisition successful nan foundry arena, Gartner expert Gaurav Gupta told The Register. The "integrated instrumentality shaper (IDM) business is very different from foundry. You person to person customer orientation, nan work [mindset], nan IP portfolio, each those customer relationships, and truthful on."
Buying Tower was a measurement for Intel to reside galore of these shortcomings, but unluckily for Intel CEO Pat Gelsinger, Chinese regulators had different ideas. Faced pinch lengthy delays Intel this week admitted defeat, paying Tower $353 cardinal to terminate nan deal.
Despite nan grounded acquisition, Gupta doesn't expect it to person immoderate awesome effect connected IFS's buildout. "Yes, it would person brought that ecosystem acquisition and expertise, but leading-edge foundry business and competing against TSMC is simply a very different ballgame," he opined.
No going back
Since announcing its entrance into statement spot manufacturing successful 2021, Intel has committed tens of billions to nan project, brought connected backstage equity partners, and lobbied for a sizable chunk of some US and EU subsidies.
To date, nan chipmaker has fabs worthy much than $70 cardinal nether building successful Arizona, Ohio, and Germany. It has besides announced billions much successful upgrades to retrofit galore of its existing accommodation for statement manufacturing of civilization silicon designs.
I deliberation Intel's champion stake is to get 1 of nan hyperscalers to perpetrate to them
So, while Tower would person played a cardinal domiciled successful advancing Intel's foundry buildout, nan spot elephantine was by nary intends limited connected it, Gupta emphasized.
And ultimately, Intel whitethorn beryllium amended disconnected focusing connected leading-edge process tech. Intel has been stuck connected 10nm process tech for years, but is moving to bring respective much precocious leading-edge nodes to market. These see nan long-delayed 7nm process – which it has taken to calling Intel 3 – and a 2nm node it is calling Intel 18A.
"From nan foundry business perspective, obviously, nan margins are much connected nan starring edge, which is wherever Intel plays," Gupta said – adding that, moreover if an acquisition is disconnected nan table, there's nary logic that Intel can't leverage its precocious packaging tech to merge different fabs' chiplets pinch its own.
It's worthy noting that while nan manufacture is moving toward a heterogeneous chiplet ecosystem, manufacture veterans show america it's still a few years away from reality. So, while partnering pinch different foundries is surely imaginable successful nan agelong term, it whitethorn not thief Intel successful nan contiguous future.
A rocky roadworthy ahead
That's not to opportunity IFS isn't without challenges. While Intel has been talking up its precocious packaging and process tech for years now, Gupta says it's clip "to move beyond conscionable having nan correct technology, to proving that it's a viable action for customers."
That intends winning customers – large ones. "They person announced a fewer customers for illustration Ericsson [and] MediaTek for Intel 18A [and Intel 16], but still, they request to do much to get immoderate of those customers up," Gupta explained.
- Biden management restricts US finance successful tech China's subject mightiness employ
- Beijing's silent curen topples Tower Semiconductor merger pinch Intel
- TSMC and pals spot successful for €10B German fab
- Gelsinger: Intel should get much CHIPS Act backing than rivals
Part of nan problem, he argued, is that galore of nan manufacturers that devour immense quantities of leading-edge semiconductor capacity besides compete pinch Intel. We fishy this could make contracting pinch Intel for manufacturing dubious for chipmakers for illustration AMD and Nvidia, particularly if they tally into challenges successful production.
Instead of courting its rivals, Gupta suggests Intel could find occurrence building chips for nan awesome hyperscalers, galore of which are processing civilization silicon pinch nan thief of TSMC. "I deliberation Intel's champion stake is to get 1 of nan hyperscalers to perpetrate to them," he stated.
An extremity to semiconductor M&A
More concerning for nan broader semiconductor manufacture is nan anticipation that Intel's grounded M&A bid whitethorn not beryllium nan past successful nan biz.
Over nan past fewer years, respective ample chipmakers person utilized their financial position to grow their IP portfolios and addressable markets done mergers and acquisitions. AMD paid $49 cardinal for Xilinx; Nvidia spent $6.9 cardinal connected Mellanox and attempted and grounded to bargain Arm for $66 billion.
But acquisitions of this standard often require regulatory support successful aggregate regions – astir notably China.
Worsening waste and acquisition relations betwixt nan US and China – fueled successful nary mini portion by sanctions connected waste and export of semiconductors, intelligence property, and AI accelerators – could time off M&A activity unfastened to retaliation by regulators.
"You don't want to put into something, do each nan owed diligence, hold for 18–24 months, and past person nan woody not spell through," Gupta said. Intel's $353 cardinal payout to extremity nan takeover must person been a bitter pill for Gelsinger to swallow.
Whether retaliation complete nan Biden administration's latest restrictions had thing to do pinch why Intel's Tower acquisition fell isolated isn't clear, but nan consequence whitethorn good person a chilling effect connected nan manufacture each nan same. ®