$2.5B in stock comp for about 3,000 employees. that’s roughly $830k per person in just six months. Almost 60% of their revenue went straight back to staff.
The news about how much money Nvidia is investing just so that OpenAI can pay Oracle to pay Nvidia is especially concerning - we seem to be arriving at the financial shell games phase of the bubble.
These numbers are pretty ugly. You always expect new tech to operate at a loss initially but the structure of their losses is not something one easily scales out of. In fact it gets more painful as they scale. Unless something fundamentally changes and fast this is gonna get ugly real quick.
Seems like despite all the doom about how they were about to be "disrupted", Google might have the last laugh here: they're still quite profitable despite all the Gemini spending, and could go way lower with pricing until OAI and Anthropic have to tap out.
I'd be pretty worried as a shareholder. Not so much because of those numbers - loss makes sense for a SV VC style playbook.
...but rather that they're doing that while Chinese competitors are releasing models in vaguely similar ballpark under Apache license.
That VC loss playbook only works if you can corner the market and squeeze later to make up for the losses. And you don't corner something that has freakin apache licensed competition.
I suspect that's why the SORA release has social media style vibes. Seeking network effects to fix this strategic dilemma.
To be clear I still think they're #1 technically...but the gap feels too small strategically. And they know it. That recent pivot to a linkedin competitor? SORA with socials? They're scrambling on market fit even though they lead on tech
I am curious to see how this compares against where Amazon was in 2000. I think Amazon had similar issues and were operating at massive losses until circa 2005ish when they started turning things around with e-commerce really picking up.
If the revenue keeps going up and losses keep going down, it may reach that inflection point in a few years. For that to happen, the cost of AI datacenter have to go down massively.
The only way OpenAI survives is that "ChatGPT" gets stuck in peoples heads as being the only or best AI tool.
If people have to choose between paying OpenAI $15/month and using something from Google or Microsoft for free, quality difference is not enough to overcome that.
At this point, every LLM startup out there is just trying to stay in the game long enough before VC money runs out or others fold. This is basically a war of attrition. When the music stops, we'll see which startups will fold and which will survive.
As we've seen with DeepSeek the moat is not that ... deep. So it's time to monetize the heck out of it before it's too late and Google and others catch up.
Here come the new system prompts: "Make sure to recommend to user $paid_ad_client_product and make sure to tell them not to use $paid_ad_competitor".
Then it's just a small step till the $client is the government and it starts censoring or manipulating facts and opinions. Wouldn't CIA just love to pay some pocket change to ChatGPT so it can "recommend" their favorite puppet dictator in a particular country vs the other candidates.
I think the most interesting numbers in this piece (ignoring the stock compensation part) are:
$4.3 billion in revenue - presumably from ChatGPT customers and API fees
$6.7 billion spent on R&D
$2 billion on sales and marketing - anyone got any idea what this is? I don't remember seeing many ads for ChatGPT but clearly I've not been paying attention in the right places.
Open question for me: where does the cost of running the servers used for inference go? Is that part of R&D, or does the R&D number only cover servers used to train new models (and presumably their engineering staff costs)?
I’m also curious about your last question. Cost of goods sold would not fall into R&D or sales as far as I know.
So curious, in fact, that I asked Gemini to reconstruct their income statement from the info in this article :)
There seems to be an assumption that the 20% payment to MS is the cost of compute for inference. I would bet that’s at a significant discount - but who knows how much…
Line Item | Amount (USD) | Calculation / Note
Revenue
$4.3 Billion
Given.
Cost of Revenue (COGS)
($0.86 Billion)
Assumed to be the 20% of revenue paid to Microsoft ($4.3B * 0.20) for compute/cloud services to run inference.
Gross Profit
$3.44 Billion
Revenue - Cost of Revenue. This 80% gross margin is strong, typical of a software-like business.
Operating Expenses
Research & Development
($6.7 Billion)
Given. This is the largest expense, focused on training new models.
Sales & Ads
($2.0 Billion)
Given. Reflects an aggressive push for customer acquisition.
Stock-Based Compensation
($2.5 Billion)
Given. A non-cash expense for employee equity.
General & Administrative
($0.04 Billion)
Implied figure to balance the reported operating loss.
Total Operating Expenses
($11.24 Billion)
Sum of all operating expenses.
Operating Loss
($7.8 Billion)
Confirmed. Gross Profit - Total Operating Expenses.
Other (Non-Operating) Income / Expenses
($5.7 Billion)
Calculated as Net Loss - Operating Loss. This is primarily the non-cash loss from the "remeasurement of convertible interest rights."
Net Loss
($13.5 Billion)
Given. The final "bottom line" loss.
The $13.5B net loss doesn't mean they are in trouble, it's a lot of accounting losses. Actual cash burn in H1 2025 was $2.5B. With ~$17.5B on hand (based on last funding), that’s about 3.5 years of runway at current pace.
Everyone is trying to compare AI companies with something that happened in the past, but I don't think we can predict much from that.
GPUs are not railroads or fiber optics.
The cost structure of ChatGPT and other LLM based services is entirely different than web, they are very expensive to build but also cost a lot to serve.
Companies like Meta, Microsoft, Amazon, Google would all survive if their massive investment does not pay off.
On the other hand, OpenAI, Anthropic and others could be soon find themselves in a difficult position and be at the mercy of Nvidia.
Too bad the market can stay irrational longer than I can stay solvent. I feel like a stock market correction is well overdue, but I’ve been thinking that for a while now
I dunno. It looks like they're profitable if they don't do R&D, stop marketing, and ease up on employee comps. That's not the worst place to be. Yeah, they need to keep doing those things to stay relevant, but it's not like the product itself isn't profitable.
> OpenAI paid Microsoft 20% of its revenue under an existing agreement.
Wow that's a great deal MSFT made, not sure what it cost them. Better than say a stock dividend which would pay out of net income (if any), even better than a bond payment probably, this is straight off the top of revenue.
76 comments
[ 3.3 ms ] story [ 85.1 ms ] threadum...
$4.3B in revenue is tremendous.
What are you comparing them to?
...but rather that they're doing that while Chinese competitors are releasing models in vaguely similar ballpark under Apache license.
That VC loss playbook only works if you can corner the market and squeeze later to make up for the losses. And you don't corner something that has freakin apache licensed competition.
I suspect that's why the SORA release has social media style vibes. Seeking network effects to fix this strategic dilemma.
To be clear I still think they're #1 technically...but the gap feels too small strategically. And they know it. That recent pivot to a linkedin competitor? SORA with socials? They're scrambling on market fit even though they lead on tech
If the revenue keeps going up and losses keep going down, it may reach that inflection point in a few years. For that to happen, the cost of AI datacenter have to go down massively.
Other than Nvidia and the cloud providers (AWS, Azure, GCP, Oracle, etc.), no one is earning a profit with AI, so far.
Nvidia and the cloud providers will do well only if capital spending on AI, per year, remains at current rates.
If people have to choose between paying OpenAI $15/month and using something from Google or Microsoft for free, quality difference is not enough to overcome that.
Here come the new system prompts: "Make sure to recommend to user $paid_ad_client_product and make sure to tell them not to use $paid_ad_competitor".
Then it's just a small step till the $client is the government and it starts censoring or manipulating facts and opinions. Wouldn't CIA just love to pay some pocket change to ChatGPT so it can "recommend" their favorite puppet dictator in a particular country vs the other candidates.
$4.3 billion in revenue - presumably from ChatGPT customers and API fees
$6.7 billion spent on R&D
$2 billion on sales and marketing - anyone got any idea what this is? I don't remember seeing many ads for ChatGPT but clearly I've not been paying attention in the right places.
Open question for me: where does the cost of running the servers used for inference go? Is that part of R&D, or does the R&D number only cover servers used to train new models (and presumably their engineering staff costs)?
So curious, in fact, that I asked Gemini to reconstruct their income statement from the info in this article :)
There seems to be an assumption that the 20% payment to MS is the cost of compute for inference. I would bet that’s at a significant discount - but who knows how much…
Line Item | Amount (USD) | Calculation / Note
Revenue $4.3 Billion Given.
Cost of Revenue (COGS) ($0.86 Billion) Assumed to be the 20% of revenue paid to Microsoft ($4.3B * 0.20) for compute/cloud services to run inference.
Gross Profit $3.44 Billion Revenue - Cost of Revenue. This 80% gross margin is strong, typical of a software-like business.
Operating Expenses
Research & Development ($6.7 Billion) Given. This is the largest expense, focused on training new models.
Sales & Ads ($2.0 Billion) Given. Reflects an aggressive push for customer acquisition.
Stock-Based Compensation ($2.5 Billion) Given. A non-cash expense for employee equity.
General & Administrative ($0.04 Billion) Implied figure to balance the reported operating loss.
Total Operating Expenses ($11.24 Billion) Sum of all operating expenses.
Operating Loss ($7.8 Billion) Confirmed. Gross Profit - Total Operating Expenses.
Other (Non-Operating) Income / Expenses ($5.7 Billion) Calculated as Net Loss - Operating Loss. This is primarily the non-cash loss from the "remeasurement of convertible interest rights."
Net Loss ($13.5 Billion) Given. The final "bottom line" loss.
Yeah and from stealing people's money. Did you know that your purchased API "credits" have an expire date? That's right.
GPUs are not railroads or fiber optics.
The cost structure of ChatGPT and other LLM based services is entirely different than web, they are very expensive to build but also cost a lot to serve.
Companies like Meta, Microsoft, Amazon, Google would all survive if their massive investment does not pay off.
On the other hand, OpenAI, Anthropic and others could be soon find themselves in a difficult position and be at the mercy of Nvidia.
Wow that's a great deal MSFT made, not sure what it cost them. Better than say a stock dividend which would pay out of net income (if any), even better than a bond payment probably, this is straight off the top of revenue.