SpaceX is a rocket company. It also became an AI company the day their founder Mr. Musk folded in xAI – inheriting a frontier model in Grok, the planet-scale Colossus training cluster, and the live data feed of the X platform (formerly Twitter). The IPO headlines fixated on the trillionaire status, the stock price and the impending growth of the company. But the question that should interest an enterprise buyer isn’t whether the stock pops, it’s whether it will offer enterprise AI that they can use to fulfill their AI ambitions.
SpaceX AI Now Looks More Like a Hyperscaler Than a Model Company
SpaceX’s AI arm has some incredible assets – enough to make it a formidable player for enterprise AI because:
It owns its compute, like the hyperscalers do. The Colossus data center in Memphis runs 555,000 Nvidia GPUs, pushing toward a million and Microsoft, Google, and AWS have similar large-scale AI data centers. The pure labs like OpenAI and Anthropic don’t; they lease their big iron from those hyperscalers.
It trains its own frontier model — but so does everyone. Gemini, Nova, Microsoft’s new MAI line, GPT and Claude, DeepSeek, Qwen, and Sakana – a competitive frontier model is table stakes now, and SpaceX has one in Grok.
It has the firehose of X posts to train models. Hundreds of millions a day – a live corpus most builders can only license or scrape. A genuine asset, but hardly unique: Meta and TikTok sit on social-data troves as deep or even deeper.
No other hyperscaler has rockets, yet. Google wants orbital compute too — its Project Suncatcher would loft TPUs on satellites, free of grid queues and permitting — but with no launch capability of its own. It’s reportedly in talks to ride up on SpaceX’s. AWS doesn’t have rockets yet – but does have a satellite program called project Leo. And Jeff Bezos just announced plans for the first time to take on external investments into Blue Origin, a rocket company he founded in 2000. Clearly this has a natural affinity with AWS. That’s a 2030s and beyond proposition at today’s launch costs, but if it arrives, SpaceX’s rivals may well reach orbit on SpaceX’s rockets.
What SpaceX Would Need to Win Enterprises
SpaceX may not even want the enterprise directly — it’s focus is consumers, government, and moonshots, not corporate procurement. But the enterprise prize is a platform to build and run agents across operations, support, and software development. If SpaceX came straight at it, here are the gaps to close:
A vision for the AI-run company. Winning enterprises takes a bold view of how a company runs on agents – like Jack Dorsey and Roelof Botha’s “From Hierarchy to Intelligence,” which imagines agents replacing the management layer. Whether SpaceX can articulate something that compelling is unclear.
A platform to build and run agents. A real platform needs both halves: design-time builder tools, and a production runtime to run agents at scale – Forrester’s Agentic Runtime Architecture (ARA), a stack of orchestration, context, tooling, governance, and observability. The hyperscalers have a head start on that runtime; xAI brings little beyond the model and some build tools.
An enterprise go-to-market. Closing Fortune 500 deals takes account teams, solution engineers, and channel partners – SpaceX sold rockets to a handful of government buyers, not software to thousands of companies. The realistic move is to borrow a salesforce through system integrators and enterprise vendors that bring real sales and services.
Trust that it’s committed to the enterprise. Enterprises sign multiyear bets and need a vendor that will stick with them – stable roadmaps, durable support, no risk of being orphaned. A hard sell for a company whose founder would rather colonize Mars and build superintelligence than serve corporate IT for a decade.
Does Buying Cursor Close the Gap?
Four days after the IPO, SpaceX moved to buy Cursor-maker Anysphere for $60 billion, the largest acquisition of a venture-backed startup on record. It’s a direct shot at the enterprise, and it closes some of these gaps more than others:
Go-to-market – the biggest win. Cursor arrives with roughly 50,000 enterprise clients and therefore developers using Cursor in those enterprises, a bottom-up wedge into accounts no SpaceX salesforce could open cold. But developer-led adoption isn’t the top-down sales-and-services motion that closes platform deals, so it’s a foot in the door, not the whole house.
The build side – partly filled. Cursor is a credible design-time surface, and with Cloud Agents one that already runs autonomous coding agents. But it’s pointed at software development, not the broad agent platform enterprises want, and it still isn’t the production runtime, the ARA describes.
Vision and commitment – unmoved, or worse. A coding tool doesn’t supply a vision for the AI-run company, and a $60 billion deal cuts both ways on trust. It signals SpaceX is serious about enterprise software, but bolts a fast-growing company onto an xAI still rebuilding after all eleven of its co-founders walked.
What It Means
Enterprise technology leaders should watch SpaceX to see the future of everything – chips, compute, orbital data centers, frontier models, and now Cursor, converging into the substrate the digital and physical world may run on. SpaceX won’t compete directly with hyperscalers for broad cloud capabilities. But watch it as a signal, not a supplier: if SpaceX doesn’t build the salesforce, services, and runtime an enterprise agent platform demands, it will not become your agentic AI vendor.
We shall see.
And of course, we can talk. Forrester clients with questions related to this can book an inquiry or guidance session with me.
Thank you to Lee Sustar, Octavio Garcia, Naveen Chhabra, and Sudha Maheshwari for their feedback on this blog.
-1024x683.jpg "Strait Outta Hormuz: Getting the Iran Oil Story Straight")
