Thread regarding Intel Corp. layoffs

Intel sold 49% of Fab24 to Apollo in 2024 for $11.2 billion. It is now buying that same stake back for $14.2 billion. This is adding 6B debt, and to the accumulated total debt of 50B.

@dq Elon seems like a good connection, to help Intel catch all possible downside when the AI buildout is revised downward to what can actually be built.

Currently only a small fraction of the forecasted datacenters are under construction. Something like 15% or so, with everything existing only on powerpoint.

I don't doubt everything will be built, but the whole exercise is constrained by all manner of physical limitations.

Add to that how the large corporations are increasingly perp-walking AI adoption, due to the need to keep from ruining their business, and this all points to a fairly radical reset of expectations.

The US is said to already be in a recession, based on various measures. Add in the supply shock from the conflict, which will create demand destruction, and that will result in corporations further moderating AI adoption (except where it is able to create cost savings by replacing human labor).

I'd expect AI to come roaring out of the economic slowdown that is taking shape, typical of any new technology adoption. LBT is enabling the company to be at least marginally more relevant to AI and computing, but there may be a deep retest in the near future, and Intel would seem to be prime for (massive) AI related layoffs.

@dr Check the stats of how many datacenter projects are able to be completed, and in what timeframe.

All the growth projections are based on a false assumption that the physical requirements can grow at any rate possible, yet they are severely constrained.

It isn't just the electrical supply, but even the transformers have a multiyear lead time, so building their own generation capacity does not speed up the project.

It isn't just corporate adoption that will naturally constrain AI, but the datacenter construction itself.

Looks like Intel will get maybe $1B a year in packaging contracts from Google and Amazon, but like the CPU sales, that is also tied to the actual, realized growth rate.

It isn't what they imagined.

Best and smartest Business leaders of the industry! Three cheers for Intel! We are ki-ling it guys!

Here is a lengthy explanation of what seems likely to be unfolding.

If it works out like this, then Intel will find itself in an existential crises, imo.

Many other companies will also be doing mass layoffs, hoping that they can adopt AI fast enough to make up the difference.

Everyone is focused on how AI may disrupt their employment, but the real risk isn't that it is capable of doing so, but rather that companies will be forced to rush into AI in hopes that they can get back to profitability.

All this at the same time as humanoid robotics are ramping furiously.

https://x.com/radigancarter/status/2035073252134129757?s=20

How silly is Intel, LOL.

It sold Fab 24, then market bo-m comes and Intel missed the capacity.

It sold Dalian memory fab to SK Hynix in 2025. Right after, memory market boomed to rocket high, Intel missed the chance badly. Now this fab is producing high capacity for SK Hynix at 30% of its revenue.

Who made these silly decisions?

@dz
A lot of people are misinformed about this AI race, everyone thinks they're racing to fully monetize the technology. It's actually a race to be the first with Artificial General Intelligence. The winner will lock down the technology and then have vast control over the entire industry. Most economists already know the investment doesn't equate to ROI in the form of paying for an AI service since the vast majority of customers will fall into a temporary use category.

@a1 Step back and look at the stock price since 2020. It is a fan shape and the next low should be in the upper single digits to mid teens.

Intel was, is and will be the semi industry canary, to tell when the oxygen has run out, at least until they can get enough external customers to support IFS.

If what is coming is as severe as the 2000-2002 period, then the company will be shutting down fabs and doing mass layoffs. This will be to survive as a going concern, not because they want to.

Rather than sell off every fab that can/will never serve an external customer (now), they will wait till they are forced to do so by the market, then sell those fabs for a fraction of the current value.

A better CFO would be making different choices. Dave was/is incompetent, and shows no signs of getting better at these strategic choices.

@dr That seems about the same as what I've read about OpenAI.

It in no way justifies the current level of forecasted Capex from the large tech companies.

This is the basic issue. AI is growing rapidly, but if more datacenter capacity is built than can be used within the first year, the economics don't work due to how rapidly the servers depreciate and how fast the technology cycles are moving.

Both hardware and software are in very fast cycles of creative destruction. The recent Google advancement which reduces a portion of the memory requirement is an example. With China pushing as hard as the US is pushing on this, the race to make AI ever-more efficient is not going to stop for some time to come.

That virtually guarantees that the hyperscalers will end up with stranded investments, because innovation is outpacing the ability of such extreme capital deployment to keep up. There was a recent case where OpenAI abandoned an Oracle datacenter project because it did not have the latest Nvidia servers.

The soon to be outdated servers will find other uses, but at much lower levels of profitability. This will eventually settle out as the scale of end users gets large enough to smooth out the range of datacenter capabilities. But there will be a serious rough patch before then, much like what happened in the late 1990s internet buildout.

@dj The revenue for AI is increasing exponentially due to coding and agents. The entire software industry and white collar workforce is being consumed. Anthropic ARR increased $6 billion in the month of March alone. It's all about the rate of change of the rate of change until it stops.

Anthropic revenue (annualized run rate):

January 2025: ~ $1B
May 2025: ~ $3B
Mid-2025 (June/July): ~ $4B
August 2025: > $5B
October 2025: ~ $7B
End of 2025 (December): > $9B

February 2026: ~ $14B
March 2026: nearing $20B (~$19–20B reported)

@dn Keep in mind that Elon has never taken a company through a recession.

The last real recession ended in Spring 2009. Everything since then was merely a contraction, which happens every few years for what prove to be short term economic disruptions.

I suspect the SpaceX IPO will end up being compared to the Cybertruck, and if a deep, prolonged recession is what is about to happen then Elon will be lucky to stay solvent.

He has so far shown interest in Intel packaging, but seems set on using Samsung/Texas for chips. F34 was there when he chose the Taylor TX facility.

@dj Thread is getting a bit off topic, but 0% rates is a key feature of socialist and communist economies, and uniformly results in bad economic decisions.

Capital ends up being deployed for political reasons. This can lead to innovation but more often just results in squandering money on investments which are ultimately not able to be sustained without additional free money.

EVs are a perfect example of this, to this day still unable to be a profitable business. Tesla still relies heavily on tax subsidies and while Rivian makes a good vehicle, it is a terrible investment.

Political manipulation of industries is a feature of all economies, and when it works out, it can lead to transformational innovation. EVs, Alt Energy, and certainly AI are good recent examples of this.

But at some point in time they have to be able to stand alone as viable business models. AI is nowhere near that point, at risk of a sharp contraction due to the wide difference in how fast the hyperscalers feel they need to grow, and how (relatively) slow the other 493 companies in the S&P 500 are willing to adopt.

OP is right. The difference in motivations between the datacenter builders and the corporate users is leading to a reconciliation, and not one that works out well for the MegaCap Tech companies (Mag7).

Be somewhere else when that happens.

I like health care, as AI adoption and robotics are already producing huge productivity gains. This seems destined to fundamentally change the valuation for the hospital chains and their suppliers.

Maybe not a good time to Learn To Nurse?

@dh Elon wants 100 billion AI5 chips... he is not going to get that on 14a anytime soon especially competing with other companies for TSMC and Samsung. AI5 is 3nm and 2nm which Fab 34 is already capable of. Power consumption is not as important for cars and robots. If he shoots them into space he gets power for ~5x cheaper than a datacenter.

@dh Much like how Tesla was enabled by 0% interest rates, AI was enabled by the post-pandemic 0% interest rate environment.

That's why you see OpenAI and Anthropic now scrambling to do ads and other ways of raising revenue, because their refinance costs are no longer 0%. They aren't getting much revenue for all they have invested and that is not a good look as the economy slowly slides into a stagflationary environment.

I mention this because AI spending is currently the main source of growth in the economy. There is room for a fairly spectacular tech implosion in the near future.

This is what happens when the central banks hold rates down for an extended period of time. This is kind of like the 1970s and Spring 2000 had a (really ugly) baby.

@af F34 will never serve an external customers, unless it is converted to 14A or newer process.

Could happen, but it is time for Intel to shed all the older fabs which are really not suited for the massive mod expansions needed to run EUV.

Holding on to them till they run out of roadmap means they will be mothballed, a waste of value.

Instead, sell them to UMC, GF, Tower or other foundries and buy back the wafers.

That will clean up the balance sheet, reduce ongoing capital needs and help make IFS profitable. Then the company would have the option to spin it off.

Hard to see the company spinning off IFS while it still owns fabs which can not support an industry standard PDK.

As the other poster mentioned, it looks like the company is putting itself into a precarious position by refusing to shed these outdated fabs, because the economic cycle has not been repealed, but merely delayed by QE.

That means the next time the central banks lose their stranglehold on the global economy, it will be that much worse, because of all the malinvestment they have created. Entire industries have never been through a real recession.

The only customer who would want massive 3nm volume is Terafab for AI5. Intel would have to upgrade the fab for Elon. It is a good stop gap measure for bringing production back to American companies.

@a9 They are lagging nodes, but certainly have some roadmap left in them.

I think the point OP was making is that this appears to be financial engineering, where the company is taking on substantial debt, at a notable loss, just so more revenue can flow to the company.

The reason why they would make this choice does seem to be the surge in datacenter CPU sales, where they have struggled to meet demand and so have some level of price strength.

The issue is as the other described, that this is all well and good unless we are at the tippy-top of the largest semi cycle since the Dot-Com era, which appears to be the case.

This is one of those things that companies tend to do when they think there are no limits to growth, and is usually followed by the shocking discovery that growth is not infinite.

Unlike Spring 2000, this time Intel has a mountain of debt to service.
Oh, imagine their deep chagrin, as they lay off workers and shut down fabs.

Intel 4 and Intel 3 are not older nodes. They are industry leading nodes in their class and still have decades of life left, especially as they shift to GPUs.

This would be a good time to look carefully at investments, and move away from the train wreck that is coming.

That means get out of any fund that is weighted to QQQ (top tech stocks).

Maybe shift to equal weighted funds, or outright short the QQQ. I like QID and stay away from 3x ETFs because of the extreme drag caused by what it takes to get that leverage.

Large Cap growth has had several of the best years it may ever have, so a down year or 2 would not be unusual. Valuations are extreme, and to the extent they are based on AI adoption, then that has room for downside.

Add in the drag on global growth from the you-know-what, and there may be nothing the central banks can do to overcome such a supply shock.

Bull markets don't end because of valuations, but Bear markets do.

Worst case scenario is Intel potentially shutting down some of the older fabs in a bid to stay solvent, and that means mass layoffs.

The last thing Intel needs is more excess capacity, especially on older nodes.

No IFS customer will ever use those fabs, and they should be sold to some other foundry.

Then do contracts as needed to keep those wafers moving, till they off ramp.

That is smart capital. This is D-mb Dave in action.