DeepSeek claims $38 of revenue per GPU, per hour for Flash V4. That is $332,880 per GPU, per year. This matches their DeepInfer numbers: 25,600 output tokens/s per GPU sold at $0.60 per million -> $55 minus some other overhead to arrive at $38.
For the bigger Anthropic and OpenAI models (and associated higher prices) there are many unknown factors. But presumably they also work hard on optimizations, and it's not so crazy to think they had or still have a lot of wiggle room to adjust prices.
Yes, but DeepSeek went absolute goblin-mode on optimization and I would not be shocked to discover that OpenAI and Anthropic were behind on that front.
OpenAI and Anthropic have more compute and the best frontier models to throw at the optimization problems. They have published articles on these efforts. For example, Dots, and Anthropic's upcoming "persistent agent" is because of these optimizations. Both companies often publish articles detailing a lot of their technical efforts. They just get buried among the flood of other news.
That's just it: they have more compute. They have a "do more with more" lever in addition to a "do slightly less with much less" lever. DeepSeek can't win with the first lever, so they have committed everything to the second, while within the frontier labs the two levers fight for attention and frankly the first is clearly the favored child. I'm not accusing OpenAI and Anthropic of being slouches, but I am not sure that more resources + divided goals > singleminded pursuit of a singular goal.
Sonnet 5.5 is 16x more expensive at $10 per million output tokens. Kimi K3.1 is also $10 per million output tokens, and its number of active parameters is only 6.5x larger than V4.1 Flash (104B compared to 16B). These things are not directly comparable of course, but just to give an idea.
Even if inference of those models needs 16x, 32x, or 48x the number of GPUs compared to V4.1 Flash, still seems like potentially great revenue per GPU.
It's sad how far down you have to scroll to get the real answer. It's because Anthropic margins on inference are absolutely bonkers and it's easy for them to drop prices by 80% and STILL have positive free cash flow
It's all based on vibes and whatever vague lies the companies put out about their financials. Marvels of corpo-nonsense speak such as "If we ignored every single cost we have, we're quite profitable!" seems to be very popular with the AI crowd
This sentiment aged extremely poorly with Anthropic's S1 now released. Turns out the "vibes" guys were right! I expect the goalpost will shift to claims that the S1 is fabricated in a larger than Enron fraud involving both the auditors and Anthropic.
You mean the S-1 docs where they state they're operating at a GAAP of -4% (even then that's based on their best 28 days multiplied by 13), meaning losing money? You sure showed the plebs with the impressive billions of dollars of loss!
Where exactly is my strawman? Scam Altman himself has said this, if we ignore the inconvenient parts of the business that loses all that money, they'd be super-duper profitable.
Anthropic said something similar, claiming to their investors they're profitable... If you ignore partner costs and the costs of training.
And mind you this is coming from a proven pathological liar who has a massive vested interest in over exaggerating his business' viability, so doubtless the "Very profitable" part is itself a massive lie.
Gross profitability always happens before net profitability, lol. "We've reached the first but not the second" isn't a deceptive claim, especially with the caveat in the same sentence, and it's a step that every net profitable company went through at some point in its life.
The correct quote is "If you ignore the costs that will come down in the future or stay fixed as userbase scales, then we are actually quite profitable"
Which is a basic truism of pretty much all startups, and nothing noteworthy or insightful to point out.
Why would you expect training costs to come down in the future? It seems to me that the labs themselves have concluded they can't stop training due to competition, which is why they're now asking the government to step in and shut down their competitors.
How can their revenue numbers match benchmarks from their inferencing code? Are they really claiming that they sell 100% of their capacity at all times, 24/7? Wouldn't they require a lot of queued batch traffic to achieve that? Or are they defining revenue in some weird way where what they mean is potential revenue?
I assume this is idealized somehow, given how close it is to DeepInfer numbers at 100% utilization. My point was more that there is seemingly a lot of margin to absorb underutilization, outages and other costs.
Given that they open sourced their inference framework, how realistic would be to buy hardware and become a provider on OpenRouter? These numbers are incredible.
Based on translating their Chinese text on Firefox, it says "per card":
> The current service throughput of a single cluster is 155 trillion tokens per day, and the overall revenue is $38 per hour (at the lowest price) per card.
With Kagi translate (best), it says GPU
> A single cluster currently serves 155 trillion tokens per day, generating overall revenue of $38 per GPU per hour (at off-peak electricity rates).
For me it's hard to imagine not starting on the second PR after submitting the first one, regardless of AI or stacked PRs. For focused work I find it important to avoid context switching, stay in the zone and keep everything in my head. Wasted work in comparison feels less significant to me.
If review is fast I'd be switching back and forth between tasks throughout the day. If review is slow I might end up implementing a feature over the course of two weeks rather than two days. With reviewers in different time zones, limited time or doing their own focused work, it adds a lot of latency going back and forth for every PR rather than iterating on a stack of PRs.
Personally I find the biggest benefit is that it lets both the author and reviewers work at the pace that works for them, with reduced context switching and latency.
For the bigger Anthropic and OpenAI models (and associated higher prices) there are many unknown factors. But presumably they also work hard on optimizations, and it's not so crazy to think they had or still have a lot of wiggle room to adjust prices.
reply