
Introducing GPU Cluster Auctions — a new way to source reserved GPU capacity on Fluence ⚡ You post what you need, providers bid, and compare available capacity and pricing in one place. Live now with 6,000+ GPUs Clusters available across the globe 🌐

Save up to 85% on AI compute. Cloud platform, GPU clusters at auction, from data centers worldwide. Powered by $FLT: fluence.ai

Introducing GPU Cluster Auctions — a new way to source reserved GPU capacity on Fluence ⚡ You post what you need, providers bid, and compare available capacity and pricing in one place. Live now with 6,000+ GPUs Clusters available across the globe 🌐

We built Fortytwo Engine to compete on intelligence per dollar, but most inference providers we see compete only on price and speed, without revealing how much accuracy gets lost in those optimizations. Thanks @fluence_project for inviting us to write about these issues
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

Open models and independent GPU clouds are widening access to AI. Confidential computing makes that capacity usable for sensitive workloads. @fluence_project’s report highlights where Phala adds value: bringing verifiable privacy into the economics of AI.
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

We asked the teams building AI compute where the GPU market is going. Four of their answers, from The AI Compute Race Report: @Theta_Network (@mitchliu) on putting idle GPUs to work @ionet (@Gaurav_ionet) on getting more from the GPUs already built @gaib_ai (@konyk001) on lending against the GPU itself @BoundlessHQ (@sshankar) on what a token is worth Their full answers are in the report 👇
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

30 to 45 weeks to get new chips. About 95% of enterprise GPU capacity sitting idle. AI has an access problem. @ionet is changing the equation. 35M+ compute hours delivered. 1M+ added every month. No lock-in. No contract tiers. Thanks to @fluence_project for including us in their new report.
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

17 of 22 GPU models got more expensive this year. The H100 is up as much as 30%. The fix isn’t only more chips. It’s putting the ones that already exist to work, just like Theta does. Proud to contribute to @fluence_project’s 2026 AI Compute Race report.
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

Picking an inference provider starts with comparing prices. If you’re running sensitive workloads, you also want to know what you can verify about how they’re handled. Glad @near_ai brought that perspective to the AI Compute Race Report 🤝
Everyone assumes private AI costs more. @fluence_project checked. In The AI Compute Race Report, private inference often comes in under the median provider. Our section covers how NEAR AI Cloud runs open models in TEEs, with attestation you can check yourself. Read the full report: fluence.ai/compute-race-2026…

Everyone assumes private AI costs more. @fluence_project checked. In The AI Compute Race Report, private inference often comes in under the median provider. Our section covers how NEAR AI Cloud runs open models in TEEs, with attestation you can check yourself. Read the full report: fluence.ai/compute-race-2026…
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

The full report has the rest: GPU prices by chip, what private inference costs, how lenders finance GPUs, and a map of 130+ companies. Read it here: fluence.ai/compute-race-2026

Lenders now price GPUs like bonds: Lambda borrowed $1.008B at 6.78% on a Baa1 rating, while brokers quote 10–15% without investment-grade customers. CME, ICE and Nodal have announced GPU futures.

For 4 of the 5 open models we priced, running the model inside a hardware-verified enclave costs less than the median standard provider. Companies can keep their prompts private on rented GPUs.

Companies on OpenRouter switched to open models. Chinese labs' share of the tokens they run went from about 10% in February to about 60% in mid-September.

Hyperscalers don’t even agree with each other: the spread between their own prices for the same chip is almost as wide as the 2x gap to GPU clouds.


The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇

Heading to @NVIDIAGTC DC? Come join us at Blackwell Evening on Dec 1🍸 🥃 👉 luma.com/0axn7b35 We're getting GPU buyers, sellers, and operators into one room for drinks, food, and honest conversations about capacity. Our team will be there all evening, so bring your questions on current pricing and availability for B200, B300, and GB300 clusters. Doors open at 5:30 PM, and we'll keep the bar and food going until 11 PM. We're a 10-minute walk from the main venue, and you don't need a GTC badge to get in.