Is It Time to Treat U.S.-Made Chips as Critical Infrastructure?

Intel introduced a first-of-its-kind co-investment program for its semiconductor fabs. The new funding model aims to help ease its gargantuan capital spending burden by sharing it with outside investors.

To that end, the Santa Clara, California-based company said it hammered out a $30 billion partnership with Brookfield Asset Management, one of the largest global asset managers, to help fund its fab expansion plans. Under the terms of the deal, Intel is responsible for 51% of the cost of building two new fabs in Chandler, Arizona, and it will maintain a controlling stake in the new facilities.

The deal stipulates that Brookfield will provide the remaining funding for the new chip-making facilities, said Intel chief financial officer David Zinsner. The companies will reportedly split revenue produced by the fabs

Brookfield, which has more than $750 billion in assets under management, touts itself as one of the world’s largest infrastructure investors, striking similar deals in sectors such as energy, transportation, and telecom.

Intel hopes the Brookfield deal will be the first of many such agreements, giving it a larger pool of capital to expand its manufacturing capacity—one of the pillars of its “IDM 2.0” strategy launched by CEO Pat Gelsinger last year.

Immense Expense

Modern chip-making is one of the most complicated and capital-intensive endeavors in the world.

A state-of-the-art fab requires roughly $5 billion for advanced analog fabs, to $20 billion for advanced logic and memory fabs, to pay for everything from the building to the equipment that gets loaded inside. According to a Boston Consulting Group report, this is more than it costs to build a state-of-the-art aircraft carrier or a nuclear-power plant.

Intel last year announced the construction of two new fabs Ocotillo campus in Chandler, Arizona, marking it as a $20 billion expansion set to be wrapped up by 2024. But the impacts of inflation, ongoing supply-chain woes, and increased equipment costs could push the price tag up to about $30 billion, warned Zinsner.

The fabs will produce chips using Intel’s upcoming Intel 20A process technology. It will overhaul how chips are built, using a whole new type of transistor called the RibbonFET, replacing the FinFET for the first time.

To help expand its chip-making capacity, Intel has pledged more than $100 billion to build new fabs within the last year and a half alone at various sites around the world, including other locales in the U.S. and Europe.

The company also has been stung by delays in upgrading its manufacturing technology. As a result, it has opened the door to the likes of AMD and Qualcomm at a time when it’s struggling with falling demand for chips more broadly.

“We have gotten behind, and that requires a fairly aggressive investment cycle over the next few years, which is not a place Intel typically finds itself,” Zinsner said on a conference call with investors and analysts.

In addition to creating more manufacturing capacity for Intel-designed chips, it is also building up its foundry services unit to make chips based on other companies’ blueprints—à la rivals GlobalFoundries and TSMC.

“Smart Capital” Plan

The agreement with Brookfield is one of several steps Intel has taken as part of what it calls its “Smart Capital” approach, which entails building fabs in such a way to adjust quickly to opportunities in the market. 

Intel is giving itself more flexibility by building empty factories or “shells” without manufacturing equipment inside at first. Then, it will decide when to install semiconductor gear depending on the markets, customers, and product readiness. Upgrading these shells into chip fabs is faster than building everything from scratch, said Zinsner.

“We’re aggressively investing in shell space to give ourselves flexibility in how and when we bring additional capacity online,” he said. “Then we equip that space, which is the vast majority of cost, in a very disciplined way as needed based on customer demand.”

Zisner noted that the deal with Brookfield is set to close by the end of this year. He is confident that Intel will strike similar deals under its semiconductor co-investment program (SCIP) as it builds more factories.

He said one of the benefits of working with Brookfield is it could enhance the company’s balance sheet by allowing Intel to tap into a new pool of capital while protecting its cash and debt capacity for future investments. The move will also preserve cash to help Intel continue paying dividends to its investors.

The company said it hopes to use co-investments, government subsidies, and pre-payments for its foundry capacity to cover about 10% of its capital spending budget, which is estimated to rise to $23 billion in 2022.

Intel eventually wants the percentage to grow to 20% or 30%.

Age of Infrastructure

The move is also further proof that the years-long chip shortage that has roiled the global economy has changed the world’s perspective on chips. Political leaders in the U.S. and Europe have started to view chip factories as a class of critical infrastructure—indispensable both to national security and the economy.

Gelsinger and other industry executives have said that they anticipate annual chip sales to hit $1 trillion by 2030 after topping $500 billion for the first time last year.

Specifically, U.S. politicians have ramped up their ambitions to rebuild America’s manufacturing prowess in chips. Earlier this month, President Joe Biden signed the CHIPS and Science Act that Intel and its peers lobbied aggressively for, allocating more than $50 billion in subsidies to help restore America’s chip-making crown and boost U.S. competitiveness with China. The U.S. is responsible for only 12% of the world’s chip output.

Intel said the co-investment it’s pursuing with Brookfield is a “force multiplier” for government incentives.

The company is not alone in having aggressive investment plans. Market research firm IC Insights forecasts that the world’s semiconductor makers will invest a record $185 billion to expand production in 2022, up 21% from $154 billion in 2021.

Intel is also increasingly seeing fabs as infrastructure that it can use instead of infrastructure it needs to build itself.

While still manufacturing most of its chips in-house, Zinsner said the company will continue to contract outside factories, such as TSMC, to produce some of its products or the chiplets inside them.

“We’ll exercise external foundry capacity to increase flexibility, optimize our roadmap, and manage our capacity,” he said. “As we move to a disaggregated roadmap, we have even more options as we can make capacity decisions on a tile-by-tile basis based on both performance requirements and the capital envelope we want to derive.”