Nvidia is enhancing its claim to be the picks and shovels of the metaverse and AI with the launch of its next-generation chips and services, that once again move it ahead of its competition and build barriers to entry.
Top of the list was the announcement of a new GPU architecture called Hopper and alongside that a new chip that uses Hopper called the H100. The take home message from this new chip is that it offers between 3-6x better performance than the Ampere A100 architecture and chip that it replaces. This has been achieved with a new transformer engine, a new processor core design as well as new formats which are designed to work optimally together.
This is another example of the trend led by Apple of wringing performance benefits through the tight integration of hardware and software.
Benchmarks are one thing, and it remains to be seen how well this new chip performs in real-world workloads, but I expect that demand for it is going to be high. This is because, in the absence of a real breakthrough in making AI more generally intelligent, improvements in AI performance are dependent on crunching more and more data and using bigger and bigger networks.
This suits Nvidia’s business model perfectly and hence, I expect this new architecture and chip to be enthusiastically received.
Nvidia is also advancing its proposition for the metaverse with the launch of Omniverse Cloud as well as an offering for industrial-scale digital twins. Omniverse Cloud makes collaboration between remote creators easier as well as making it easier to access processing-intensive tasks from remote locations.
The idea between industrial digital twins is to create a completely realistic version of an asset like a football stadium or factory and be able to monitor it in real-time or use it to run simulations to see how the asset will react to certain situations. I think that the metaverse is certainly going to grow in the enterprise first given how bad the consumer user experience remains and the productivity improvements that can be attained even now.
These two new offerings round out the Omniverse offering and add weight to Nvidia’s ambitions to be the operating system of the metaverse. Nvidia does make the right noises when it comes to the metaverse as its proposition claims to enable any other metaverse to connect to it, which is what RFM research has concluded will be crucial if the metaverse is going to take off.
The net result is that Nvidia is justifying its position as the leader in the provider of AI silicon and clearly has no intention of letting this go without a fight. I think that the company has a lead in this area that is going to take some time to overcome and sets a high bar for Intel to overcome.
However, even with the recent correction, the shares are already pricing in a lot of benefits from these launches hitting the bottom line. This is why although I like the company, I continue to struggle to see how the current multiples attributed to its earnings are justified.
This is why I prefer Qualcomm or MediaTek, which offer slightly lower growth but at a much more reasonable price multiple. The risk/reward balance is still not in Intel’s favour and I think that the shares can still go lower before the current risks are priced in.