Nvidia’s bad day

At PGM, we actively manage risk today while targeting outperformance tomorrow.

So no matter what investment risks concern you most, from geopolitics to inflation to liquidity, PGEM brings disciplined risk management expertise that spans 30 market cycles.

Our active approach finds opportunities and volatility, helping our clients to navigate risk and achieve their long-term goals.

PGEM, our investments shape tomorrow today.

Pushkin

The last thing we need right now is some kind of shock to the US tech sector.

And guess what?

That's exactly what we've got.

Mad scenes in the markets this week after Nvidia, the Mahusive US-listed chip maker, said it would take a $5.5 billion hit to earnings because of new controls on sales to China.

Its stock dropped 7% on Wednesday.

That's a lot for a company worth $2.5 trillion-ish dollars.

That's a trillion with a T.

Today on the show, we're asking, is it all over for US Tech?

What's going on here?

This is Unhedged, the Markets and Finance podcast from the Financial Times and Pushkin.

I'm Katie Martin, a markets columnist here at FT Towers in sunny London, locked and loaded for a four-day weekend.

Hooray!

Joining me down the line from New York City, I have a dynamic duo.

It's Aiden Writer from the Unhedge newsletter, but also special guest, John Foley, head of Lex and low-key tech nerd.

Guys, thanks for joining.

Thanks for having me.

There you go.

So listeners, if you get confused, John is the Brit with just a hint of an accent from, I believe it's Hull.

Am I correct, John?

It is Hull.

Yeah.

Uh nerd, it's gonna snur.

That's what they say about that.

This will confuse the Americans.

So, John, you have been sold to me as tech man.

Tell me what is going on with NVIDIA.

Well, NVIDIA's stock fell this week quite dramatically, wiped off almost $200 billion,

billion with a B of market capitalization, because they said that they had discovered that the chips that they sell to China, which facilitate AI, are going to be subject to a strict new licensing requirement, which means that they may not be able to sell them to China at all.

So, they're taking a five and a bit billion dollar accounting hit, and analysts are kind of scrambling to work out what this means for their revenue and for the growth generally.

China's not a huge part of what NVIDIA does.

It's maybe about like round numbers, like 15% of revenue, a bit less.

But it's important, it's an important customer.

And of course, the chips that NVIDIA, just to backtrack a bit, like NVIDIA is the company that makes the chips that make AI possible.

Like everyone who's anyone in artificial intelligence is basically reliant on NVIDIA's microprocessors.

Just stepping back a little bit, these were chips that NVIDIA had already specifically designed for the Chinese market, is that right?

Right.

So because there have been restrictions on what you can sell to China for quite a long time, NVIDIA cannot sell its most high-end chips to China.

So it created this range called the H20 that was compatible with the rules as they then were.

So China's been buying.

It's the H20.

I thought it was H2O.

No, it showed what I know.

Oh, wait, maybe it...

No, I'm sure it's H20.

I'm sure it's H20.

Because

they have

the 200.

Let's call it the H20.

People can write in, if not.

So it's the H20, and China had been importing these chips to do what it, you know, it's like do what you can with what you have.

But now the US government is saying that even those are not acceptable.

And this is partly a response to what we call the deep seek shock from early this year when suddenly China turned out to have created this large language model, an AI model, that was actually way better and seemed to have been made on a shoestring, way better than anyone thought.

So that kind of sent the willies through governments that are trying to make sure that they win the AI race.

And so now here we are where NVIDIA may not, still might, but it may not be able to sell these chips anymore into the People's Republic of China.

Yeah, and crucially it was made with those weaker chips like the H20 that were compatible with US export controls.

Well, yes.

So it seems that DeepSeek was made with export rule compatible chips and not that many of them as well, crucially.

However, there is separately a probe going on at the moment in Congress to find out whether China has been able to get access to chips that it wasn't supposed to.

And NVIDIA has been drawn into that because they're asking for

the House of Representatives Committee on China.

They're asking NVIDIA to give more information so they can establish whether these export controls are being sidestepped somehow.

So what's the deal with

these rules and restrictions?

Is it part of the tariffs conversation, or is it completely separate?

It's separate, really.

This is the thing with chips.

There are two prongs that are pinching this industry from different sides.

One is, yeah, tariffs.

These are physical goods that go backwards and forwards across borders.

And also they make up things that go backwards and forwards multiple times.

So So, if you have tariffs, every time something crosses a border, these things can get really expensive.

But the export controls are actually different, they're more of a security issue because AI is a national security issue for everyone for multiple reasons.

One is that it's just a big economic driver that promotes lots of activity, it's going to create efficiency, supposedly, and change business models in national companies.

So, everyone wants to make sure that they're capturing as much as they can of that activity.

But also, potentially, don't forget that AI is a weapon.

So, if anyone reaches the kind of commanding heights of AI, what we call artificial general intelligence, where you've got AI that is as intelligent or more intelligent than humans, then that potentially becomes a very powerful geopolitical tool.

So, America doesn't want to see China get that first, China doesn't want to see America get that first, it becomes basically a new space race.

And these things are kind of separate, but they're happening at the same time.

And that sucks for you if you are, for example, NVIDIA or any of the other chip makers that are reliant on free trade and also the ability to sell the products they want to the customers who want them.

Gotcha.

It also sucks for you if you're an investor, right?

Because, you know, there was a period a few months ago when pretty much the only thing that we talked about on this podcast, it felt like at times, was NVIDIA.

Because it came from like relative obscurity to become suddenly a $3 trillion company making enormous amounts of money every quarter, just churning out the profits, absolutely just dominating the world.

And it became the biggest company in the world, right?

And it occupied this enormous slice of the US stock market.

But that means it got itself to such a size that when it takes a hit like this, the whole market feels it.

So, Nasdaq was down.

That's the very sort of tech-heavy US index.

That was down pretty heavily yesterday, which was Wednesday.

Was it down about 3% or something?

The SP 500, which is more of a kind of general index, that was down by a little bit less, but still a lot.

Some of the indices that are really focused on tech, like the Philadelphia Semiconductor Index, how much is that down?

Like 20 something percent this year.

So yeah, I guess I'm wondering, you know, tech has had such a good run.

It's been so dominant.

It's like led the entire US and global market by the nose.

What's the sense over there?

You know, is this starting to crack a little bit?

How widespread is the stress in tech?

It depends how far you look back.

So if you think about what we call the Trump bump, you know, Trump gets into office, stocks kind of roar ahead and then sag this year.

And tech has sagged, as tech tends to do, tends to move with the market and then some.

So tech stocks, I had a look this morning and the big US tech stocks, so anything over a billion dollars, on average down 10% since the US election, but were up very strongly at one point.

NVIDIA is down more than that.

NVIDIA's down 25%.

AMD, which is another chip maker, the sort of closest rival to NVIDIA, I guess, but still a very distant one, is down much more than that, like 37%, 38%.

So tech is always more volatile, but if you look back further, these companies are still worth way more than they used to be.

So as you said, Katie, NVIDIA, you know, five years ago, no one was talking about NVIDIA.

In the not that distant past, NVIDIA just made chips for video game, basically graphics.

And then it was discovered, not by NVIDIA, actually, that these things also were really good for AI calculations.

So the companies now were worth 15 times what it was five years ago, even now after that correction.

That's craziness.

Some of what we've seen this year has been about the actual AI narrative, right?

So

will it pan out is the bigger question, but you know, DeepSeek changed how these companies think about what they need to invest in, whether that's big data centers versus that models, et cetera.

But we've also just had a huge risk-off then in the US, and it seems like tech sold off the most.

In your mind, is that a reflection of things changing in the AI space and the tech space?

Or is that people just locking in the gains after those huge run-ups you mentioned?

This is actually the really interesting question for me because it could be either I think that what's happening is just that there is a kind of market vibe shift.

For example, you take what happened to NVIDIA with these export controls.

That doesn't really affect the rest of the tech sector at this stage, but you see valuations weaken because people are just a bit twitchy and tech is a volatile industry.

And so there is a kind of people, like the cliche, the sort of hackneyed phrase is like animal spirits, I guess, right?

But animal spirits are dampened a bit.

People are just feeling a bit less, you know, feisty or whatever.

So they're locking in their winnings, basically.

And also, they're locking in their winnings, and there are ample winnings here, right?

So there's that.

But the question of whether AI is fundamentally going to be different from what we thought it was going to be, whether the demand trajectory is going to change,

where the future returns from all this investment in data centers and large language models, whether that's going to change, I don't think that has happened yet.

I don't think people have recalibrated their expectations of what the returns from AI are going to be, which is a risk because at the moment we've got the tariff crisis or the tariff adjustment or whatever you want to call it.

We haven't yet had the had-let's title.

Let's

call it a crisis.

We haven't yet had the AI crisis, and the AI crisis is where people is not where people say, like, AI is not a thing, it's where they say, Oh dear, we've invested way too much too soon.

And that if we get both of those crises together, that will be a very big event for the stock market and for

preserve us.

Well, I mean, that's what I was going to ask, right?

Because there's lots of people who've been kind of jumping up and down on the sidelines of markets for a really long time saying

AI is overhyped, AI is overvalued, most importantly from our point of view.

And there's been a lot of people saying, you know, just you wait.

One of these days it's going to be just like the dot-com crash of, you know, 2000, 2001, where, you know, this whole thing is going to fall over.

What do you think are the possibilities that we are at the, you know, at the foothills of something really ugly like that?

Or do you just think, look, AI is real and that's just rubbish?

What I'm going to say is the greatest hits of things that other people have said because this has been so well discussed.

That's what we do for a living, John.

Don't tell them that.

Here's how I see it.

So is it the dot-com bubble?

I mean, not really.

Like, the dot-com bubble had a lot of business models that just made literally no sense, right?

There was just no plan to make any money ever.

It was the pets.coms and so on.

Like, a comparison that's often made instead is kind of with the telecoms bubble or the telecom infrastructure bubble, which sort of happened around the same time where there were lots of inflated valuations, but lots of money went into stuff that actually did turn out to be highly useful.

Like the internet, it turns out, has been very useful for us.

I use it quite often.

Yeah, I do as well now and again.

So you have wasted money, but you also just have redistribution of value.

So a company may have invested in something and that company may no longer be here, but we're still benefiting from its efforts.

So I think AI is more like that.

I don't think AI is going away.

I just think

we don't really understand exactly what it can do or what kind of impact it's going to have on society and on companies.

But clearly, the answer is it's going to have a big one.

It's just a question of whether the specific businesses that are most highly valued right now are the ones that win and are still here later.

I mean, think about something like OpenAI, right?

OpenAI is the iconic artificial intelligence company.

It's made a lot of money.

Its valuation has gone very high.

It's certainly advanced AI tremendously.

It's created some amazing things.

Chat GPT started this whole hype cycle.

That's part of the OpenAI stable, right?

Yeah, ChatGPT, sorry, is the product through which you and I basically access OpenAI's AI technology.

Like, will OpenAI be the winner?

Will it still be here in its current form 10 years from now?

Like, who knows?

Maybe, maybe not.

But I think the things that ChatGPT is doing for us will absolutely still be here.

I mean, it seems like from DeepSeek, the thing that really changed, maybe the people who got hurt the most were people like OpenAI, who are the model creators, right?

DeepSeek kind of showed that anybody can make a model or that it is easy to make a specialized model for individual companies.

It's not that expensive.

What you do need is something to build it upon.

So they're always going to be there in that regard.

But it seems like open AI is now more of a commodity within this space.

And the big implementers, so whoever can actually implement it, whether that be meta, whether that be someone else, is going to profit.

I mean, is that your take of how DeepSeek has affected some of this?

I think that the big impact from DeepSeek, and speaking not as like I'm not an AI nerd in the sense that I don't

spend my hours going through different models and trying to work out exactly which one is better than which other one.

I do all the time.

I I was going to defer to you on this case because that's what you are of course known for.

So I think what DeepSeek did was it was the like good enough moment.

So it was the moment when people were like, you know, actually, like it's great that Open AI and Google, for example,

Perplexity, whoever is going for these like extremely ambitious goals.

But actually, if I'm a business, I just kind of want something that's good enough.

And if it's cheap, great.

And that's where the applications are.

That's where the actual usage is going to be.

And that's where the revenue is going to be.

Already companies are using AI, and companies are making money from AI, and they're using it for fairly mundane, routine things, they're using it for their call centers, they're using it to pull their disparate data together a bit more effectively.

And you don't necessarily need to be the absolute best ever cutting-edge AI model to give people what they want in that sense.

DeepSeek was the moment where I think a lot of the innovators here were like, actually, it turns out if you're good enough, that's fine for 90% of people who are going to pay you money.

Well, the big picture here is, you know, if it's good enough and it's Chinese, then fine, we'll use it.

This all plugs into the really kind of big contentious issue of the day, which is, do we need super shiny American technology?

Do we even want super shiny American technology?

So I was talking to a banker from an American bank over in the UK the other day.

And she was saying that they were talking to a prospective European client who was saying, well, I might be interested in signing up with you guys, but I want you to keep my data in Europe or in the UK.

I don't want you snaffling my data away and putting it in the States because who knows what might happen to it over there.

So I think there is a bit of a sort of vague cloud of a question mark that's hanging over US tech to say, hang on, this is sufficiently strategic that maybe we do need some European tech champions, some Asian tech champions, so that we can do this stuff ourselves and not be so beholden to these pesky Americans.

Is that a risk?

So there are two sides to it.

One is the question of

sort of geopolitical-linked risk.

Do you want to have all of your precious data being managed by a company that is not part of the US government, but may be

in some ways susceptible to influence from the US government?

This is like the old argument for Chinese companies, right?

Alibaba say,

which is a big cloud computing provider in China, like it's not state-owned, but does it have to do what the government tells it to?

Like effectively, probably yes.

That was never a question with the US really, but now it kind of is a bit, because we've seen American companies doing in some cases things that we previously thought they wouldn't in order to keep the government happy.

So if you're a European business owner, then yes, you are kind of thinking about that.

That, though, might get a bit over blown, overstated at some times.

I mean, while it would be lovely if every country had their own big tech giant, we've seen that Europe has struggled to get any tech company off the ground for a long time, right?

One thing that I always come back to with that sort of thing though, Aiden, is

this is what you Americans always say, right?

There's no alternative to US tech.

There's no alternative to the US dollar.

There's no alternative to US government bonds.

You guys need to like open your minds.

There's so many people around the world who really don't like you very much at the moment and they're trying to figure out how to do these important things themselves.

So, I don't know.

I mean, I don't disagree, but you know, there is some pretty intense tech regulation in the EU that some would argue is stopping investment.

Also, again, we don't have a big EU tech company that is competing with these Americans.

So, at some point, it feels like there's something structural that has inhibited that from happening.

One thing to also remember is that Europe doesn't have a stock market big enough to host a company of the size of Nvidia or Google.

It has like a billion different stock markets that don't really talk to each other.

Mario Draghi over here.

We got to get the unified capital markets going.

Exactly.

I'm just saying you Americans need to be humble and act normal.

That's my number one message to Americans.

Okay, listen.

Let's stop being mean to Americans and come back in a sec with Longshore.

Bonds are back.

And so is All the Credit, P Jim Fixed Incomes Monthly Podcast Series.

From the latest trends to long-term perspectives, you'll get timely fixed income insights from leading economists, research analysts, and investment professionals.

Whether you're new to bonds or a seasoned investor, tune in to All the Credit wherever you get your podcasts.

This podcast is intended solely for professional investor use.

Past performance is not a guarantee of future results.

Oak it oak, it's time for long short, that part of the show where we go long a thing we love or short a thing we hate.

Who should I pick on first?

I'm going to pick on Aiden first.

I am long Fed volatility.

Fed chair Jay Powell came out this week and really hit home what I think he's been trying to say for a while, that we don't know what the outcome from tariffs will be, and it's possible that will endanger the Fed's current mission to keep prices under control and make sure the U.S.

economy is still growing at a reasonable rate.

As one might expect, President Donald Trump did not like that and hit it back, and he said he's very excited for Chair Powell's, quote, termination.

I think it's pretty clear that he knows he can't fire Jay Powell, but I do believe there's going to be a lot of missives, you know, broadsides and other attacks over the Fed, and they're going to use their power to appoint people who might not be as non-partisan as we're used to.

So I think we're in for a bumpy ride in the next year and a half until Chair Powell's term ends.

He has a seriously difficult job.

John, son of Hull, what you got?

So, sir,

I'm going to preface this by saying that I'm not telling you that you should actually short this company's shares.

Well, this is investment advice, people.

Conceptually, I would short ASNL, which is an enormous European maker of machines that make chips.

They etch chips.

Their machines are amazing.

They do things like fire a laser at a moving droplet of molten tin and hit it twice in order to etch chips.

So they're really good at at that.

However, both their customers seem to be delaying orders.

We found that this week, and their shares fell dramatically on that, which links to all the AI things we've talked about before.

Also, they are a company that makes stuff, ships it across borders, adds stuff to it, ships it back, etc., etc.

So, in a world of tariffs, they get hit multiple times on all of their enormous, expensive machines.

So, I think they are one to watch because they're kind of at the heart of both the AI hype cycle and the tariff crisis.

Good one.

Mine is

different.

I am limit short sperm racing.

I don't know if you saw this.

There was a story this week in the Times, the British Times, not the New York Times.

Oh, turns my stomach.

Let me tell you.

I'll quote, a startup called Sperm Racing, run by four teenage entrepreneurs from the US, it's always you lot, says it's raised $1.5 million to stage the event at a Hollywood Palladium on April 25th.

Gonna pit samples taken from two healthy young university students against each other on a racetrack 20 centimetres long and modelled on the female reproductive system.

It is an enormous no from me.

There's nothing good going on here.

Please join me in being short.

Sperm racing.

I'm so short.

I mean, I'm short, but I kind of want to find out who wins.

Oh, Aiden, the only thing is that someone on the internet quipped that it should be called Spermula One.

I quite like that, but everything else about it is bad and

I'm not into it.

So on that cheerful note to send you on your way for the long weekend, guys, the UK is not in on Monday.

The US markets will be open with no adult supervision, so please don't break them.

Just please.

I'm all out of patience with you lot.

Listeners we will be back in your ears on Tuesday so listen up then.

Unhedged is produced by Jake Harper and edited by Brian Erstad.

Our executive producer is Jacob Goldstein.

We had additional help from Topa 40.

Cheryl Brumley is the FT's global head of audio.

Special thanks to Laura Clark, Alistair Mackey, Greta Cohn and Natalie Sadler.

FT Premium subscribers can get the Unhedged newsletter for free.

A 30-day free trial is available to everyone else.

Just go to ft.com/slash unhedged offer.

I'm Katie Martin.

Thanks for listening.