Another week, another addition to my SIPP — the pension pot I treat as the 'adventurous' corner of my portfolio. This week I opened a small position in CoreWeave (CRWV), an AI cloud infrastructure company that went public recently. Like my other SIPP holdings, this is speculative, the position is tiny, and I'm giving it decades to play out inside a tax wrapper. If it works, brilliant. If it doesn't, the damage is contained.
A quick reminder on how I approach my SIPP. The vast majority of the money sits in broad ETFs — S&P 500, global all-cap, a bit of FTSE 100. That's the engine. The individual shares are the satellites — small, high-conviction bets on themes I believe have genuine structural tailwinds. Collectively, these individual picks make up less than 10% of the SIPP. I'm not gambling the pension on any single idea. Each position is kept deliberately small.
So what is CoreWeave and why did I buy it? At its simplest, CoreWeave builds and operates GPU-accelerated data centres. They buy thousands of Nvidia GPUs, build the infrastructure around them, and rent that compute capacity out to AI companies who need it. Think of it as the cloud computing layer specifically designed for AI workloads — training large language models, running inference, rendering, and scientific computing. They don't build AI models themselves. They provide the picks and shovels that everyone else needs.
What makes CoreWeave interesting — and what got my attention — is their relationship with Microsoft. Microsoft is their anchor tenant, committing to long-term contracts that provide a base level of revenue that most early-stage cloud companies can only dream of. That's not a guarantee of success — contracts can be renegotiated, relationships can change — but having one of the biggest AI spenders on the planet as your core customer provides a level of demand visibility that's rare for a company this young.
The scale of what they're building is genuinely significant. They operate a fleet of tens of thousands of Nvidia GPUs across multiple data centres in the US and Europe, with a roadmap to scale far beyond that. Their infrastructure is purpose-built for AI — high-bandwidth networking, liquid cooling, optimised storage — the kind of specialised setup that general-purpose cloud providers weren't originally designed for. This focus is their edge, though whether it's a durable competitive advantage remains to be seen.
The company went public recently, which means it's early days as a listed entity. Early-stage companies carry specific risks that more established firms don't: limited trading history, less analyst coverage, lock-up expirations that can create selling pressure, and the possibility that early growth rates don't translate into steady-state profitability. CoreWeave's revenue is growing at an extraordinary rate — powered almost entirely by AI infrastructure demand — but with that growth comes significant capital investment requirements.
The risks are real and I want to be upfront about them. This is a company burning through cash to build out infrastructure, with massive capex plans ahead. Their customer concentration is high — Microsoft represents a significant portion of revenue, and while that's a strength today, it's a vulnerability if the relationship ever changes. Competition from hyperscalers building their own AI-specific infrastructure is a legitimate concern. And the valuation reflects enormous growth expectations — if AI demand moderates or shifts, the stock would be vulnerable.
That's precisely why the position is small. I'm not making a large bet on CoreWeave succeeding. I'm making a small bet that the AI infrastructure build-out — the physical data centres, the GPUs, the networking, the power — represents a genuine multi-decade investment theme, and that CoreWeave is positioned to capture a meaningful slice of it. If I'm right, the small position grows into something more significant. If I'm wrong, I lose a tiny amount of capital and move on.
One thing I've learned over the years is that the most valuable companies in any technology wave aren't always the ones making the headlines. During the internet build-out, the big winners included the companies providing the infrastructure — the networking equipment, the servers, the fibre optic cable. During the cloud computing revolution, AWS became a profit engine for Amazon. The AI era will have its own infrastructure winners, and I'm trying to identify them early while keeping my bets appropriately sized.
CoreWeave joins my growing collection of AI-adjacent SIPP holdings. Constellation Energy provides the nuclear power for AI data centres. Nebius Group provides AI cloud infrastructure with Nvidia's backing and Nasdaq-100 inclusion. Cipher Digital is turning from Bitcoin mining into AI data centre real estate. And now CoreWeave — pure GPU cloud for AI workloads, with Microsoft as its anchor. They're all connected by the same theme: the massive infrastructure build-out that AI requires, which I believe is underappreciated relative to the attention lavished on AI models and chatbots.
Are these sensible investments for someone my age? By conventional standards, probably not. But they're small, they're in a pension I won't access for years, and they're backed by structural trends rather than short-term speculation. I'm not recommending anyone follow me. These are speculative positions — all of them could decline significantly. The majority of my money remains in broad ETFs because that's what I have the most conviction in. These satellites are interesting, exciting, and kept firmly in their place.
As always: this is what I did with my own money. It's not a recommendation. Do your own research. All investing carries risk — you can lose money, and past performance doesn't predict future returns. These are particularly speculative positions — concentrated, volatile, early-stage, and not suitable for most investors. I'm comfortable with the risk because the positions are tiny relative to my overall portfolio and sit inside a pension I won't access for years. Your circumstances, goals, and risk tolerance are different from mine.