Intel is said to be in talks to buy GlobalFoundries in a deal worth as much as $30 billion dollars. This would be an enormous acquisition for Intel if the rumor is true. As always, such rumors should be taken with a grain of salt.
The Walt Street Journal is reporting this particular rumor, which comes as GlobalFoundries is planning an IPO. The WSJ notes that whatever Intel is considering, “talks don’t appear to include GlobalFoundries itself as a spokeswoman for the company said it isn’t in discussions with Intel.”
The WSJ goes on to explicitly note that GlobalFoundries might continue with its plans for an IPO, which raises the question of whether these talks are happening or not. One way to make a company look good before an initial public offering is to float rumors of a takeover. Intel has pledged to invest tens of billions of dollars in chip manufacturing since Pat Gelsinger returned to the company as CEO, but buying GlobalFoundries would represent a large additional commitment to the idea of Intel Foundry 2.0.
Let’s assume, for the sake of argument, that the rumors of these talks are real. There are several reasons why Intel might buy GlobalFoundries and several why it might not. On the positive side of things, buying GlobalFoundries instantly gives Intel a foundry customer base and a group of employees who are experienced in running a near leading-edge foundry. GlobalFoundries may not be on the leading edge any longer, but the company’s array of 12nm, 14nm, and 28nm product lines are in hot demand given the global semiconductor shortage.
Buying GF also gives Intel immediate access to foundry facilities that are already tuned for manufacturing for others, as opposed to being designed for Intel silicon built according to Intel design rules. There’s nothing stopping Intel from either overhauling one of its own existing facilities to serve as a client foundry, but overhauls and new construction both take time.
The acquisition would also give Intel some credibility as a foundry provider rather than a manufacturer for its own products first and foremost. When AMD spun GlobalFoundries off into its own company, the firm acquired Chartered shortly thereafter for much the same reason.
As for why Intel wouldn’t buy GlobalFoundries: Much depends on what kind of customers Intel plans to target and how it plans to target them. Intel could build a client foundry business that caters to high-performance customers on the leading edge who want custom process nodes. It could build a business focused on providing low-cost hardware in bulk on older nodes. This type of activity is more often associated with second-tier foundries like UMC, SMIC, and GlobalFoundries than with TSMC or Samsung, but legacy foundries continue to exist because manufacturing on older nodes can be quite profitable in volume.
It can be challenging to mesh corporate cultures during an acquisition and any equipment transfers/installations also take time. It may not be clearly cheaper or faster to buy GF than to build a new facility depending on the type of customers Intel wants to pursue and the benefits it wants to offer.
Buying GF would make Intel and AMD foundry partners, for as long as AMD needs to purchase I/O die and legacy 14nm GPUs. Even this would arguably play to Intel’s advantage. What better way to emphasize the difference between Old Intel and New Intel than to fab for AMD? If Intel wanted to take advantage of another foundry’s position to challenge TSMC, GlobalFoundries is the company they’d probably pick on the basis of size and relative market strength.