1,252%. That’s the revenue jump Nscale reported for the first half of 2026 as it filed to go public in the United States. Not 12%. Not 125%. Four digits, and the decimal point isn’t a typo.
My job is picking apart AI tools and the companies behind them, and I’ve learned that a number that big usually tells you less than it appears to. So before anyone starts sketching out what NSCL looks like on a chart, let’s talk about what a figure like that actually means, and what the same filing quietly admits alongside it.
Big percentages are a math trick, not a moat
Percentage growth is the most flattering statistic in tech, and it gets more flattering the smaller you started. A company going from a modest base to a decent base can post numbers that look like a typo. A company going from huge to slightly more huge posts single digits and gets called mature. Same business quality, wildly different headline.
Nscale didn’t disclose in the material I’ve seen what the starting base was, and I’m not going to guess at it. What I can say is that in AI infrastructure, this pattern is common. You sign a couple of large capacity deals, revenue recognition kicks in, and your year-over-year comparison becomes theatre. It isn’t dishonest. It just isn’t the same thing as a durable business, and treating the two as interchangeable is how retail investors end up holding the bag.
The net loss is the part that matters
Nscale posted a net loss. It’s also reportedly aiming for a multibillion-dollar valuation. Those two facts sitting in the same filing are not a contradiction in this market, they’re basically the template.
Building AI cloud capacity is brutally capital-heavy. You buy GPUs, you buy or lease data centre space, you pay for power, you pay for cooling, and you pay all of it before a single customer invoice clears. That’s why growth and losses scale together here. The interesting question isn’t whether a company like this loses money. It’s whether the losses are the cost of building something that eventually prints cash, or the cost of renting out expensive hardware at thin margins while hoping demand never cools.
From the outside, with the facts currently public, I can’t tell you which one Nscale is. Anyone who claims they can from a 1,252% headline is selling something.
Being Nvidia-backed cuts both ways
Nvidia’s involvement is the detail that will do the most work in headlines, and it deserves a bit of skepticism rather than reverence. On the positive side, backing from the company that makes the chips you need is genuinely useful. Supply access in this market is a real advantage, and Nvidia doesn’t put its name on infrastructure plays casually.
The flip side is concentration. When the dominant chip supplier is also an investor in the companies buying those chips, the whole structure gets circular. Demand for Nvidia silicon looks stronger partly because Nvidia helped fund the buyers. That doesn’t make any individual business bad. It does mean “Nvidia-backed” is a supply-chain fact, not a quality rating, and I’d rather people read it that way.
What I’d want to see before forming a view
If I were evaluating this the way I evaluate an AI product, I’d be looking past the growth rate entirely and asking about:
- Customer concentration. How much of that revenue comes from the top handful of contracts? A 1,252% increase driven by two customers is a very different risk profile than one driven by two hundred.
- Contract length and terms. Multi-year committed capacity is a real asset. Short-term burst usage that migrates to whoever is cheapest next quarter is not.
- Gross margin, separate from net loss. Losing money while building is fine. Losing money on the underlying service is a structural problem.
- Power and site access. In AI infrastructure, electricity and land are becoming the actual bottleneck, not chips.
- How much capital the business still needs. The IPO is one round. It is rarely the last one.
None of that is in the headline. Most of it will be somewhere in the filing, and that’s where anyone with actual money on the line should be reading.
The honest take
Nscale is going public into a market that is unusually willing to pay for AI exposure, and it’s bringing a number designed to travel well. That’s smart timing, not a character flaw. British infrastructure companies listing on the NYSE under a four-letter ticker in this environment is about as rational a decision as exists.
My caution isn’t about Nscale specifically. It’s about the reflex that treats a four-digit growth figure as a verdict. AI infrastructure demand right now is real and enormous. It’s also being financed by people with a strong interest in it continuing, and the companies riding it are mostly losing money while they build. Both things are true at once. Read the filing, not the percentage.
🕒 Published: