From MS-DOS to AI Agents: OpenAI, NVIDIA, X Reflect on Future of AI

AI Summary: Microsoft's Build 2025 conference highlighted a remarkable convergence in the future of AI, as discussions with Sam Altman (OpenAI), Jensen Huang (NVIDIA), and Elon Musk (X, Tesla, SpaceX) revealed shared visions despite their distinct approaches. Key themes included the shift towards "agentic" AI for delegated programming tasks, the exponential acceleration of computing infrastructure driven by advancements like NVIDIA's new chips, and the critical importance of grounding AI in physical reality for both capability and safety. This collaborative environment, with Microsoft acting as a central hub, underscores a period of unprecedented innovation in computing.

May 20 2025 14:57
Microsoft's Build 2025 developer conference has become tech's must-watch event, particularly for its high-profile conversations with industry leaders. This year was no exception as Satya Nadella engaged in discussions with Sam Altman of OpenAI, Jensen Huang of NVIDIA, and Elon Musk of X, Tesla, and SpaceX about the state and future of AI.

These conversations revealed a remarkable convergence of vision among three distinctly different innovators, each approaching AI from their own angle yet arriving at complementary conclusions about where computing is headed. Their insights paint a picture of a computing landscape evolving more rapidly than at any time in history.



Sam Altman's appearance centered on OpenAI's recent launch of Codex, an AI coding agent that represents a fundamental shift in how developers will approach programming. "This is one of the biggest changes to programming that I've ever seen," Altman remarked, describing a world where developers can delegate complex tasks to AI partners.

The vision Altman outlined isn't just about AI assistance but true delegation. "This idea that you now have a real virtual teammate that you can assign work to... at some point say like, I got a big idea, go off and work for a couple of days and do it," represents a fundamental shift in how software engineering will function. He described parallelized workflows where developers can simultaneously fix bugs, implement features, and get questions answered about their codebase.

What makes this shift particularly notable is that it arrives after years of incremental progress. Altman reminded viewers that the first version of Codex debuted in 2021, but the path from code completion to true agentic capabilities required significant model improvements that are only now coming to fruition.



Jensen Huang's conversation with Nadella took a different angle, focusing on the infrastructure making these AI advances possible. His insights revealed how NVIDIA and Microsoft are reimagining computing at the most fundamental levels.

"40x speed-up over Hopper. That's just an insane speed-up in just two years," Huang noted about NVIDIA's new GB200 chip compared to its predecessor. This pace dramatically outstrips traditional Moore's Law progression, creating what Nadella described as "Moore's Law on hyperdrive."

But perhaps Huang's most interesting insights concerned fleet management rather than peak performance. He advocated for continual, incremental upgrades rather than massive periodic replacements: "When technology is moving 40 times per generation and it's 40 times every two years, you really want to upgrade every year. You don't want to wait every four years."

This approach, maintaining backward compatibility while pushing hardware boundaries, creates compound benefits. Huang explained how even older hardware in the fleet dramatically improves over time: "Hopper on Hopper has been 40x over the last two years" through software optimizations like "in-flight batching, and speculative decoding, and all kinds of new algorithms."

The result is what Nadella characterized as "tokens per dollar per watt" - a new efficiency metric driving the entire industry forward.



Elon Musk's provided yet another perspective, focusing on the importance of physical grounding in AI development. Musk, who began his career programming on MS-DOS before Windows existed, brought a long-term perspective to the conversation about Grok, his AI project now launching on Azure.

"With Grok, especially with Grok 3.5 that is about to be released, it's trying to reason from first principles, so apply kind of the tools of physics to thinking," Musk explained. This physics-based approach, he argued, is essential for both capability and safety.

Musk emphasized repeatedly that AI must be "grounded in reality," noting that "physics is the law and everything else is a recommendation." This reality grounding comes naturally in his companies' other AI implementations: autonomous vehicles must navigate real roads safely, and humanoid robots must manipulate real objects correctly.

"To really be intelligent, it's got to make predictions that are in line with reality, in other words, physics," Musk argued. This grounding in physical reality provides natural guardrails for AI systems that purely language-based models might lack.


Despite their different focuses, all three conversations revealed striking alignment on key principles that will shape the next generation of computing:

• The compounding effects of hardware and software co-evolution
• The shift toward agentic, delegated computing
• The importance of physical grounding and real-world testing
• The integration of models into broader systems rather than standalone capabilities

This convergence suggests we've reached an inflection point where the theoretical becomes practical across the industry. The multi-year visions these leaders described are now materializing into products developers can actually use.

Altman emphasized the need to embrace new workflows, noting how quickly OpenAI's internal developers who adopted Codex "changed their workflow... the incredible amount they were able to do relative to someone else was quite interesting."

Huang stressed the importance of architecture stability for developer productivity, explaining that "all of our software architectures are compatible... the rich ecosystem of tools and all the developers that are developing on it, they're anxious to develop on it because the install base is so large."

Musk took the most direct approach, ending with a plea: "I can't emphasize enough that we're looking for feedback from you, the developer audience. Tell us what you want, and we'll make it happen."


Microsoft has positioned itself as the crucial nexus connecting these innovations - running OpenAI's models, deploying NVIDIA's hardware at unprecedented scale, and now hosting Musk's Grok on Azure. This platform approach allows the company to benefit from advancements across the ecosystem while providing customers access to best-of-breed technologies regardless of origin.

As Huang observed, "This is the insane execution of our two organizations now hyperlinked." The same could be said for Microsoft's relationships with OpenAI and now with Musk's xAI.

The technological trajectories outlined at Build 2025 suggest we're entering a period of accelerated innovation in computing that will reshape how software is built, deployed, and experienced. The abstractions that have defined computing for decades are being reconsidered at every level.

Altman suggested simpler model selection is coming: "The models will get simpler to use. You won't have so many to pick from. It'll just sort of automatically do the right thing." This represents a significant shift from the current landscape of carefully selected model parameters.

Huang described a future where the distinction between training and inference blurs, with dynamic fleet management allocating resources based on emerging workloads. And Musk emphasized that truth-seeking and physical grounding will become increasingly critical as these systems take on more responsibility.

What remains clear is that developers embracing these new paradigms earliest stand to gain the most. As Altman noted, "We haven't seen many technological shifts like this in history, but every time one has come, like leaning in early and hard has been very rewarded."