The contraction of AI deals, triggered by recent disappointing quarterly earnings, has been a disaster for big tech companies.
The selling pressure that began with Tesla and Google is spreading and gaining momentum across major cloud computing and semiconductor stocks, led by Intel, whose shares fell 27% today, its worst one-day drop in decades, after the company suspended dividend payments and pledged to cut 15,000 jobs on Thursday.
Sentiment worsened after the Financial Times on Friday obtained a recent letter from hedge fund Elliott Management purportedly telling investors that Nvidia and the entire big tech sector are in a “bubble” and that artificial intelligence is “overvalued.”
Nvidia has entered a bear market, dropping more than 20% from its June high, but that decline has only erased about two months of gains. As Elliott warned, Nvidia’s stock is trading well above levels seen for much of May and has more than doubled since January, suggesting that its valuation may be overvalued.
Nvidia, which traded roughly in line with the Nasdaq’s 2.7% drop, declined to comment when contacted by Fortune about the report. The company could not be reached for comment.
The worsening economic outlook hasn’t helped Nvidia or its semiconductor peers either. The Federal Reserve’s most recent overnight lending rate of 5.33% is about 230 basis points above the June consumer price index, a gauge of inflation, raising the speculation that monetary policy is too restrictive.
Amazon and Tesla considering developing their own AI training chips
In a sign of the new trend of selling stocks when they are rising, semiconductor design company Arm Holdings Inc. reported a 39% increase in quarterly revenue on Wednesday but lost a quarter of its market capitalization this week alone. Its growth forecast only matched analysts’ expectations, a sign the market got too ahead of itself.
Essentially, little has changed at Nvidia. The company’s AI training and inference chips remain popular for their unparalleled ability to process the terabytes of data needed for generative AI. In fact, they may be too good: CEO Jensen Huang has struggled to keep up with the pace of demand, leading tech companies to invest in their own microchips as a kind of Plan B.
Elon Musk said last week that Tesla will double down on its investment in the Dojo chip because the company is keen for Huang’s H100 processor to power a data center it plans to build in Austin. Optimized for training neural networks on video data rather than text, the chip will give Tesla the ability to design its own autonomous driving solutions that Nvidia can’t.
Similarly, Amazon CEO Andy Jassy said the company is funding its own chips.
“We have a close partnership with Nvidia,” Jassy told investors on Thursday, “but we’ve heard loud and clear from investors that they appreciate better price/performance, which is why we’ve invested in our own custom silicon: Trainium for training and Inferentia for inference.”
