Co-Founder & CEO of @JoinVantage: infrastructure cost management.
New York, NY
Joined October 2008
- Tweets26.6K
- Following602
- Followers3.7K
- Likes39.7K
Launch day: @JoinVantage now supports @CoreWeave costs.
For teams that train or serve their own models, GPU capacity is often the biggest part of the AI budget. But CoreWeave usage lived in the CoreWeave Cloud Console. Getting it next to AWS or your model providers meant exporting usage, pricing it against your contract by hand and uploading it as a Custom Provider.
Now you connect a CoreWeave organization with an API access token and we import usage every day by SKU, cluster and region. CoreWeave's FOCUS export doesn't include costs yet, so we price it with a per-SKU rate sheet that starts at CoreWeave's on-demand list prices and that you can update to your negotiated rates. Then use Virtual Tags to allocate each cluster's cost to the team that owns it.
Available today for all Vantage customers.
vantage.sh/blog/coreweave-co…
1
9
535
Launch day: @JoinVantage now supports @togethercompute costs.
Teams use Together AI for open-model serving, dedicated GPU endpoints and fine-tuning, often right next to AWS, GCP or Azure. But until now that spend lived in Together's own cost analytics, separate from the rest of the AI budget.
Now you connect a Together AI organization with an API key and we pull in daily costs by product, model, GPU, project and endpoint. Dedicated Inference is broken out by GPU type and on-demand vs reserved GPU-hours. Serverless Inference is broken out by model and input, cached input and output tokens. All of it sits next to your other AI providers and the rest of your cloud bill, with Budgets, Anomaly Detection and Virtual Tags.
Available today for all Vantage customers.
vantage.sh/blog/togetherai-c…
3
248
Picked up a raspberry pi after work, worked with Muse to write a python file to persist the data from the Aranet4...it works...and to state the obvious: AI is magical.
Setting up a regular cron job to cache the data for the long term now.
What's behind this is a long, tiresome, righteous legal battle with my apartment building that you can see I will not lose :)
1
92
👀
NEW: Apple is cutting employee AI spending budgets even as it pushes workers to use the technology more. Engineers have seen daily limits cut from $300 to $200, with further reductions to $100 planned. Other teams could see limits fall to $25. bloomberg.com/news/articles/…
4
807
We have an ever-increasing number of new Vantage accounts being created by AI agents...so we're making it a bit easier for them to get going.
We just added devoted endpoints and documentation specifically for agents to create new Vantage accounts: docs.vantage.sh/auth
4
221
Launch day: @JoinVantage now supports @DeepgramAI costs.
Voice is turning into a real AI line item. Speech-to-text, text-to-speech, voice agents. But until now Deepgram spend lived in the Deepgram Console one project at a time, totally separate from the rest of a team's AI budget.
Now you connect a Deepgram project with an API key and we pull in daily costs by model, API key and tag, right next to OpenAI, Anthropic and the rest of your cloud bill. If you serve lots of customers from one Deepgram project, tag their usage in Deepgram and you can allocate voice AI spend by customer with Virtual Tags.
Available today for all Vantage customers.
vantage.sh/blog/deepgram-cos…
2
206
Fun one: Vantage now breaks down GitHub Copilot costs by user.
Since GitHub moved Copilot to usage-based AI credits, spend moves with how much each developer uses it and which models they pick. But the bill still only shows one total per organization per day. The customers we talk to who wanted to know who was driving that total were exporting GitHub's usage reports and reconciling them against the bill and its discounts by hand.
Available today for GitHub Enterprise Cloud customers. More here: vantage.sh/blog/github-copil…
1
6
334
Another interesting Anthropic S-1 detail: Anthropic is about to pay a competitor ~$3.1B a month for compute.
Its SpaceXAI deal hits ~$3.1B/month at full delivery this December. To put that in perspective, that's more than double what Anthropic spent on all of its compute, everywhere, across the first half of 2026 (~$1.5B/month).
Total value: up to $120.5B. Paid to a rival.
Curious....is there is any precedent where so much is being paid out to a vendor who is a direct competitor?
One of the more under-appreciated details in Anthropic's S-1:
Nearly half of Anthropic's revenue now flows through cloud marketplaces - 48% in the first half of 2026, up from 11% in 2023. That's roughly $7.8B of revenue in six months sold through Bedrock, Vertex and Azure Marketplace instead of direct.
Which means the hyperscalers get paid twice! They sell Anthropic the compute - Anthropic has committed hundreds of billions of dollars to AWS, Google and Microsoft - and then they take a fee on roughly half of the revenue distributed through their marketplaces.
Also on gross margin...those marketplace fees don't sit in cost of revenue. They're in sales, marketing & partnerships, so they never touch the headline 85% gross margin everyone is quoting today. Fees paid to channel partners alone rose ~$2.7B year over year in H1 2026.
Hyperscalers feel like they're kind of the winner here all around though reading through this.
Just in case anyone was curious if you can make it in one day from metro north to NJ transit…the answer is yes.
And now, I shall eat a pizza.
3
17
863
Ben Schaechter retweeted
One answer to what's driving GPU rental prices:
Token demand on OpenRouter is up ~27x since last November, from ~6T/week to ~165T/week
Open-weight models now handle ~72% of those tokens, up from ~28% last November
Charts of the Week: a16z.news/p/charts-of-the-we…
Zooming out, GPUs are still renting ahead of a standard 6-year depreciation schedule
Rental price vs. schedule:
- H200: +143%
- B200: +111%
- RTX PRO 6000: +73%
- H100: +54%
- A100: +9%
Most of the gap opened up after April, when token demand took off
Charts of the Week: a16z.news/p/charts-of-the-we…
👀
I think you're right about this happening at the most advanced (or earlier adopting) businesses. At the same time, the % of companies who have done the full arc of tokenmaxxing to token questioning to token economics is still very small? And so, there's plenty companies who are just about to discover tokenmaxxing, which makes me think that overall spend will keep going higher.
338