The post says xAI’s latest business looks less like a frontier AI lab winning on model quality and more like a landlord for scarce AI compute. The core claim is simple: if Google and Anthropic are paying enormous sums to rent xAI capacity, then xAI may have found a real business, but it is a different business from the one implied by frontier-lab valuations. It looks more like monetizing GPUs, power access, and fast datacenter buildout than monetizing a leading model.
That framing mostly held up. People did not dispute that the rental business could be lucrative right now. Many argued the opposite of the bearish version on one narrow point: this is not “just power” or commodity colocation. The scarce assets are high-end
NVIDIA GPUs, high-bandwidth memory, permits, skilled datacenter builders, and especially power supply that is actually online. In a market where demand still exceeds supply, renting deployed compute can throw off serious cash. Several commenters said older GPUs are not behaving like normal depreciating hardware because shortages are so severe that even H100-era or consumer cards still command unusually high prices.
But the thread landed on a harder distinction. Good infrastructure economics do not rescue frontier-lab economics. If xAI can repay buildout costs quickly by leasing capacity, that supports the case that the datacenter was a smart speculative bet. It does not prove Grok is competitive enough to deserve the same multiple as OpenAI or Anthropic. A lot of people saw the lease deals as evidence that xAI could not productively use all of its own compute at the frontier, which is exactly the opposite signal you would want from an
AGI race contender.
The other major thread was financing risk. Plenty of commenters were uneasy about the web of ownership, leasing, and valuation markups across Google, Anthropic, xAI, NVIDIA, and SpaceX. The sharper point was not that every related-party or circular-looking deal is fake. It was that these arrangements can blur what is operating demand versus what is ecosystem support, and that public-market investors may price short-term scarcity revenue as if it were durable software-like growth. Several people pushed back on the laziest Enron analogies and said real token revenue exists, real demand exists, and “circular financing” is being overused as a catchall term. Even so, the consensus was that useful AI products and overextended valuations can coexist, much like the internet did during the dot-com era.
On xAI’s actual product, the mood was dismissive but not absolute. Most saw Grok as behind the top tier overall, especially for coding, yet some said it is genuinely strong for current-events questions, web search, and less refusal-heavy professional use. That gave the thread a cleaner conclusion than the headline alone: xAI may have built a very valuable compute business, but that is not the same thing as winning the model race, and investors should stop pretending those are interchangeable.