Bubbles rarely pop gradually — they tend to burst. Yet the recent cooling of AI hype has, paradoxically, delivered a welcome reprieve for Nvidia.
Shares of the AI chip giant surged 5% in after-hours trading on Wednesday after the company reported fiscal third-quarter results that eased some investor jitters. The rally provided a sharp rebound from a 10% slide over the prior three weeks, a decline fueled by mounting anxiety over the seemingly limitless capital pouring into artificial intelligence — with no slowdown in sight nearly three years after ChatGPT ignited the boom.
That unease has rippled far beyond Nvidia. The Nasdaq Composite has dropped 5% so far this month, while other AI-heavy names — Advanced Micro Devices, Oracle, and infrastructure player CoreWeave — have suffered even steeper losses. For once, a little air coming out of the AI enthusiasm bubble has actually worked in Nvidia’s favor.
Dan Gallagher for The Wall Street Journal:
All that clearly got Nvidia’s attention. “There’s been a lot of talk about an AI bubble,” Chief Executive Jensen Huang said Wednesday on the company’s earnings call—marking the first time he or any other Nvidia executive has uttered that phrase during the company’s calls, according to a transcript search by AlphaSense. “From our vantage point, we see something very different.”
That difference was marked mainly by the company’s quarterly data-center sales climbing 66% year over year and topping $50 billion for the first time. That beat Wall Street’s projections for a business segment that is now larger than what most other chip companies produce in total annual revenue. The company’s adjusted operating margin rose by a percentage point from the quarter that ended in May, despite its efforts to boost production of its technically ambitious Blackwell family of AI systems.
“The No. 1 take is they are a machine,” said Sean O’Hara, president at Pacer ETFs.
The law of large numbers hasn’t seemed to catch up to Nvidia yet. Huang last month projected $500 billion in sales for just two generations of the company’s AI chips over a two-year period ending next year. Nvidia typically refrains from projecting beyond a single quarter ahead, but it suggested Wednesday that such an aggressive target could actually prove conservative.
MacDailyNews Take: Plenty of room to run.
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!
Support MacDailyNews at no extra cost to you by using this link to shop at Amazon.
The problem is more with other companies in the AI sphere. Nvidia is making money selling hardware to the AI crowd. Look CoreWeave which is strapped for cash and now borrowing money to pay interest on other loans. Look at Oracle whose stock has dropped ~30% from recent highs. They are on the hook to provide billions of dollars in services to OpenAI which may have no money to pay them. They are stretching the rubber band tighter and tighter. We’ll see if they can generate revenues to relax this or if the situation will break.
Sure there’s room to run. Run in the opposite direction from the leather jacket snake oil salesman. You know the con is being prolonged when NVDIA is shoveling its irrational exuberance investor capital into propping up its unprofitable customers. Let me repeat that: Huang isn’t investing in better products or efficiency, he’s putting excess cash into companies that are unprofitable and have no clear path to becoming profitable in many years, if ever.
Never mind the glorious vision that the tech gorillas have, to replace humans with chatbots. That bizarre vision cannot be realized without a resulting economic contraction when tens of thousands of tech workers are let go. Of course those companies that trigger recessions will then pivot to achieve their real goal: attracting federal stimulus and corporate welfare.
You shouldn’t start believing in AI until 3 things happen:
1. the AI companies make cash instead of burning it
2. someone presents a believable plan to generate the electricity these sites suck up
3. you can actually measure a benefit that the average person can identify, like stoplights cycling efficiency or a dramatic reduction in insurance costs due to AI replacing inefficient overhead. instead AI is currently just a meme generator for lazy ad industry dorks who have a low bar for credibility or truth.
Until then …. what is the point of playing the greater fool game? NVDA is down 13% from its peak and has a lot of room to fall. it will not prop up the economy when main street falters under the increasing pressures of isolationist trade policies and the aforementioned announcements for tech layoffs.
And yes, the macroecon is softening. Home Depot already announced low sales because the middle class is tightening its belt in the current trend of persistent high inflation (nit fixed as promised). Verizon is axing thousands of jobs. Oracle’s bubble is also deflating. AI isn’t actually making these companies more profitable.
Let us know when you’re great again. Wall Street’s been delusional. NVDA isn’t essential to economic wellbeing, it’s just the latest overpriced stock on the market. Watch out when the bulls step off the elevator. The bubble will pop loudly and the bloodbath will be widespread.
You might consider investing in a company that makes real products that consumers need, rather than helping WSJ inflate and already large bubble any further.
True that. Huang claimed in investor call that his chips are “sold out”. Then a few days later, the quarterly filing showed that in reality, unsold inventory actually increased in the last quarter.
Big tech oligarchs’ promises of putting data centers in everybody’s backyard are running up hard against local opposition to subsidizing these energy hog projects, not to mention the real backlogs of energy infrastructure builders required to turn on the lights.
So while the oligarchs think they own the federal government and will sweep aside any hindrances such as pesky environmental regulations to stop their literal electricity power grab, they have no realistic plans to make the energy they want. Has anyone added up the costs to consumers to build all the renewable, gas, coal, and nuke plants necessary? Oh wait. It’s still not enough to satisfy their lyin’ asses. Nevermind coal miners are a literal dying breed, nukes aren’t cost effective due to both cost of imported unsustainable fuel AND infinite waste storage at guarded facilities, and the admin has directly and intentionally attempted to kill renewables with obscene tariffs. So, good luck slapping up gas turbines everywhere, Sillycon Valley broligarchs. Why should we pay for this when your products are nothing but erratic parrots?
“AI bubble fears are unfounded?”
That’s a leap of faith unsupported by the evidence.
Pride come before a fall.
Sorry about the length but I want to provide full context.
If you lived through the .com bubble, I was still young then but have read voraciously about it and the 2008 crisis, I think we are in the equivalent of 1998 this year and next will be the last gasp of the AI stock mania (not AI use itself) and 2027 will be when it pops.
In 1998, I dropped out of college to “cash in” (to help early pay down the student debt I was building and needing income to pay the rent etc) and worked for Wherehouse Entertainment (Wherehouse records stores HQ in the west) as a tier 3 tech. In late 1998 I was asked to attend a big meeting for a new internet venture they were doing, by an executive. He wanted me to be ready to support it, naming me 1st POC, making it equally top priority for support as the internet group (#1 priority support at the time). The pitch to all the execs was to address the problem of slow internet because most people were on slow dial up, was to open “.COM stores” where you would theoretically travel to a .COM store, equipped with high speed internet and listen to a large catalog of music and then place your orders for delivery in store. So tell me, does it make sense to drive or ride to a store to listen to music that you cannot leave with? Before the days of streaming or internet accessible music other than Napster. It made no sense to me either and when the Exec asked what I thought, I mentioned he doesn’t need me to bring on additional support help for that, I am sure me plus 1 backup will more than cover it with the network team if the internet or routers go down. He picked up that I didn’t think it would work, and said he’s considering scaling back how many .COM stores they will initially open (from about 6-10 to start down to 3 to test) “to help make sure there isn’t to much on my support plate” wink wink and I nodded. This was the example to me of thinking anything .com would succeed, today it’s anything AI is expected to succeed.
Moral of my long story is that, when it gets to everyone is SURE anything AI will succeed is when it will pop, just like in 2000 when the .com bubble popped. At least that’s my perspective.