History or hype? Either way, Nvidia gets paid
CASH IN YER AI CHIPS
Welcome to Mostly Cloudy! Today: Nvidia shows off its AI muscle and plans for cloud services, the FTC is circling cloud computing companies, and a cybersecurity legend is gone way too soon.
Nvidia CEO Jensen Huang speaks at GTC 2023. Credit: Nvidia
Nvidia CEO Jensen Huang marshalled the spotlight this week to remind everyone in tech that the road to our glorious AI-powered future rolls through Santa Clara.
At the company’s annual GTC conference on Tuesday, Huang promised that “we are at the iPhone moment of AI,” which might be a bit overstated but there’s no denying the groundswell of interest in generative AI technology spurred by OpenAI’s work over the last year. Nvidia’s A100 and H100 chips are the engine for this new wave of AI investment, and this week Huang introduced dozens of new products and services at GTC and expanded the company’s partnerships with the major cloud providers.
But perhaps the most significant introduction was DGX Cloud, a new service launching on Oracle but coming soon to other major cloud providers.
You can rent Nvidia’s AI chips from all the major cloud providers as part of their compute services, but much of the heavy lifting required to get anything done falls on your shoulders. DGX Cloud is basically a managed version of those services that comes with a lot of Nvidia’s software expertise, and for a mere $36,999 per instance per month, allows a business “to access its own AI supercomputer using a simple web browser.”
After all the hype, this could be a real inflection point for generative AI: Businesses around the world have a serious case of AI FOMO right now after months of promises that generative AI is the new technology magic bean that will finally unlock their company’s (and their competitor’s) potential, and will be under serious pressure from executives and board members to figure out an AI strategy. During this experimenal period renting AI capacity will be cheaper than buying it, especially if Nvidia manages the hard parts.
It’s also another sign that Nvidia has become the most important chip company in enterprise computing, which has been clear to investors for a while but is really starting to hit home.
Demand for basic compute services powered by Intel and AMD isn’t going to disappear any time soon, but businesses that believe they need AI capabilities will increasingly depend on Nvidia’s GPUs as the more strategic part of their tech budgets. And unlike the competitive revival we’ve seen in the server processor market, Nvidia doesn’t have a real challenger for AI workloads in the data center.
But cloud providers will likely get tired of paying Nvidia’s AI toll and develop serious rivals to its chips, which Google has been doing for years while others take steps in that direction. Until then, however, the company that stands to benefit most directly from the new AI boom is Nvidia.
Sound off in the comments
The FTC announced this week that it will be “seeking comment” on the cloud computing industry, which is good news for all those people at tech conferences who pretend to ask speakers questions after a presentation.
Specifically, the commission wants to examine “the extent to which particular segments of the economy are reliant on a small handful of cloud service providers” as well as look into “incentives providers offer customers to obtain more of their cloud services from a single provider,” among other things. “The reliance on cloud computing is only set to increase, as industries such as healthcare and finance continue to migrate to the cloud, and as computationally intensive novel technologies such as chatbots powered by large language models proliferate,” it said in a separate blog post.
FTC Chair Lina Khan helped create a 2022 report produced by the House Judiciary Committee on tech competitivness that singled out AWS, saying “evidence suggests that Amazon has also taken steps to lock in and extend this dominance in ways that
risk harming customers, businesses, and the broader public.” Proving that remedies should be enacted to do something about it, however, will be much more difficult.
Around the enterprise
Kelly “Aloria” Lum, a widely known and respected figure in cybersecurity and hacking circles, died Sunday at the age of 41. Lorenzo Franceschi-Bicchierai described how she’ll be remembered.
Software development is too expensive, and if software can be reliably produced by generative AI technologies the enterprise tech world will look very different, Paul Kedrosky and Eric Norlin of SK Ventures argued in a fairly convincing essay.
GitHub certainly wants to make that world happen: It just introduced GPT-4 to its Copilot programming assistant.
Until this week, AWS had managed to avoid the brunt of Amazon’s layoffs. That’s no longer true.
Microsoft Loop, an Asana competitor first shown off in 2021, is now available as a preview to a limited number of sign-ups.
Thanks for reading — see you next week!