Nvidia has had quite the big challenges during the past a couple of years. The company, which was long known as a maker of graphic cards, transformed into not just “one of the IT guys”, but as the biggest, most influential and most expensive company in the world.
Nvidia became a key developer and manufacturer of AI-focused processors and very expensive configurations for AI data centers. The results were quite profound. Nvidia broke several records and became the first company ever to reach $3 trillion market cap. And while everything seemed to be going amazing, August and September 2024 didn’t go all that well for the company.
All of a sudden it lost more than $400 billion of its market cap in a vast stock market selloff. Then it clawed half of that loss back, but the notions about the company and its future had already changed. Nvidia may still be a market and an (IT) industry leader, but it’s no longer the investors’ dream darling. Now they are much more careful about every move of the company.
Nvidia is now suddenly in big trouble. How did that even happen? What will be the effect on the data center and on the IT industry as a whole? Let’s explore.
“Shaming Bitcoin”
Nvidia is making billions of dollars of revenues and profits from its AI chips. Yet, in two weeks around the end of August and the beginning of September 2024, the company lost about one fifth of its value. It began September by losing a total of $406 billion. Such a sudden fall was described by the financial media as “shaming Bitcoin”, because up until now such stock market behavior was typical of cryptocurrencies. All of a sudden, not just anyone, but the top company in the world started netting massive drops and fluctuations in its stock prices.
“Nvidia’s volatile prices makes Bitcoin look like the epitome of calmness”, says Investor. The company’s 30-day volatility dwarfed Big Tech’s performance up to 80 points. It was 4 times bigger than Microsoft’s and twice as big as Bitcoin’s. It was Nvidia’s worst market performance in more than two years.
And while the size of the drop was unexpected, the actual fall in prices isn’t much of a surprise. Nvidia’s “market immunity” simply ended. The company enjoyed the investors’ euphoria who were buying the stock no matter what the news about Nvidia were. It was simply the hottest property on the market and everyone wanted a piece of the cake.
But the news kept piling. And eventually Nvidia couldn’t escape the realities anymore. First, Nvidia’s new Blackwell chips are delayed. Then, the market forecasts were lowered in general. Next, there’s a possible antitrust court date looming for the company as the US Department of Justice is looking into some concerns. Broadcom’s disappointing forecast for the market overall was the final straw.
The future is a bit foggy
The main problem for Nvidia is that investors’ scrutiny can be very difficult to overturn. Once investors lose their confidence in a stock, they usually swing to the other end and become overly pessimistic. And start to find problems in everything or simply notice details that before they overlooked in their enthusiasm (or sometimes on purpose).
One such issue is the fact that nearly half of Nvidia’s revenue currently comes from just four top clients, Fortune reported. No one knows who they are, but it is known each of them is buying Nvidia’s hardware en masse, to the figure of $3 billion and more. Each. So, more than $12 billion of Nvidia’s $30 billion revenue for the second quarter of 2024 comes from just four clients.
That’s a big red flag for the future as these companies won’t need new hardware every month of every year. They are investing big now, as they are building new AI data centers and expanding the current ones. While the names of the clients aren’t mentioned, it’s not difficult to figure out who’s big on AI lately and is building massive data centers to accommodate their projects. Once their work is done, t will be a question of gradual upgrades rather than buying as much CPUs and GPUs as possible as soon as possible. Thus Nvidia needs not four big clients, but a rotation of 400 smaller ones that would regularly need chips in order to have sustainable and predictable revenue.
That worries investors, but it’s also an opportunity for them. Goldman Sachs feels Nvidia’s stock now is a bargain exactly because AI is going beyond the hyperscalers. It makes the company’s stock more accessible and also signals exactly the moment mentioned above. The hyperscale data centers are getting close to satiation thus the market will open up for the rest.
“People thought that Nvidia was going to continue growing at that pace. The company is now in a strange situation where, even though their growth rate is still high – their revenue was up a record-breaking 15% from last quarter – it still falls short of market expectations, with shares taking a 7% hit.Practically any other CEO in the world would be overjoyed with the growth that Nvidia is seeing, but because it’s not like the obscene surges that were seen in prior years people are finding it a little bit underwhelming,” tech analyst and writer Chris Stokel-Walker says to The Guardian.
The real challenges are yet to come
So far, we’ve focused mostly on the financial side of Nvidia’s woes. That’s because on the IT front the company is the sole leader and no one can get close to the performance of its AI chips. Or maybe not. The competition is working hard to change this.
On one side we have AMD. They recently announced the new UDNA GPU architecture. It unifies RDNA and CDNA into one with the goal to take on Nvidia’s CUDA systems. “So, part of a big change at AMD is today we have a CDNA architecture for our Instinct data center GPUs and RDNA for the consumer stuff. It’s forked. Going forward, we will call it UDNA. There’ll be one unified architecture, both Instinct and client [consumer]. We’ll unify it so that it will be so much easier for developers versus today, where they have to choose and value is not improving”, says AMD’s Jack Huynh, the senior vice president and general manager of the Computing and Graphics Business Group during IFA 2024, quoted by Tom’s Hardware.
The optimization is in order to put more focus on the AI and data center side while not leaving the consumer GPUs fall behind. The shift in focus is clear – AMD wants to take on Nvidia’s key businesses.
But wait, there’s more. The California-based startup Cerebras Systems is working on the third generation of its Wafer Scale Engine which will power the Cerebra Inference solution. At its core we have a massive chip with a GPU, accompanied by massive 44GB of SRAM thus eliminating the need for external memory.
The Cerebras Inference can deliver 1800 tokens per second for Llama 3.1 8B and 450 tokens for Llama 3.1 70B. Cerebras Systems claims an average of 20 times faster performance than Nvidia’s GPUs.
The Wafer Scale Engine features about 4 trillion transistors per chip. Thus the chips are ideal for enterprises requiring fast processing of AI models. It reduces latency to the bare minimum and improves the processing of large volumes of data in real time. The chips might be a bit too focused on AI while AMD and Nvidia’s solutions are bringing more flexibility for the wider range of applications for their chips. With that said, AI is now the main focus for many and that will continue to be the case for the foreseeable future. So, Nvidia should not neglect this competitor and ready up appropriate solutions.
So, is Nvidia going back to being “one of the lads”?
One could say, that Nvidia’s days as the top dog are now over and the company will slowly go back down among the rest of the chip makers and will return the top market positions to Apple and Microsoft. Especially since the Blackwell launch was delayed.
But there’s a flip side to that coin. The demand and expectations for said Blackwell are extremely high. This is one of the main reasons for the delay. The company needed the extra time to make sure it will get things right as it fully realizes this is going to be a historic, era-defining product launch for the company. One that could change and shape the IT industry and the pace of AI development as a whole.
So, it’s too soon to say that Nvidia’s glory days are over. The company still has a lot of potential to deliver on. And there are still years ahead of constant data center construction and expansion projects. So, the need for powerful AI-capable hardware will still be here for a while. The competition will also force the company to continue to innovate and deliver on its promises. Nvidia might not deliver the historic record-breaking performances on such a short timeframe anymore, but that’s actually a good thing for the company if it manages to keep its market share and competition advantage. It would be a sign that it has matured its market and it has forged stable and lasting relationships with its clients.