A Futures Market for AI Tokens Lands $3 Million in Seed Funding

The YC-backed startup is applying commodity-trading logic to AI compute, letting providers and buyers lock in prices for future model runs.

About Touchmark AI

Published

The interface is a spreadsheet, but the asset is time. You specify a model, a volume of tokens, a cloud provider, and a delivery window sometime next quarter. The price is fixed today, a discount to the on-demand rate you’d pay if you just hit the API. You are not buying compute hours or reserving a GPU. You are buying a forward contract on AI inference, a promise of future intelligence, traded like wheat or copper [Touchmark, November 2026].

This is the core transaction inside Touchmark AI, a San Francisco startup founded earlier this year that is building a neutral marketplace for buying and selling future AI capacity. Its bet is that the explosive, unpredictable demand for running large language models will create a need for financial instruments to hedge risk and smooth supply. For application builders, it’s a way to lock in costs and guarantee capacity for a planned feature launch. For inference providers,from large clouds to specialized GPU clusters,it’s a way to monetize idle future capacity and attract committed volume [Touchmark, November 2026]. The platform, which launched commercially in August, already reports more than 30 providers listing over $30 million in inference contracts, with customer requests for quotes covering more than 4 trillion tokens [Touchmark, November 2026].

The Commodity Wedge

Touchmark’s founding insight is that AI inference is becoming a standardized, measurable commodity. A token from GPT-4o running on Azure is functionally identical to one from the same model on another compliant provider. This fungibility, the team argues, is the prerequisite for a liquid market. The platform’s product allows customers to reserve capacity with service-level agreements, consume the tokens later via API, and even rebalance or resell their commitments through a shared ledger [Touchmark, November 2026].

  • Dynamic procurement. Instead of the traditional model of flat, often wasteful GPU reservations, Touchmark matches a customer’s specific demand shape with a live, dynamically priced pool of capacity from multiple sellers [Touchmark, November 2026].
  • Price discovery. The marketplace establishes a forward price for inference, giving both buyers and sellers a benchmark beyond today’s spot rates. The company claims discounts can reach 10-30% off on-demand pricing [Y Combinator, retrieved 2026].
  • Risk management. For an AI startup planning a major product release, volatile or unavailable inference capacity is an existential operational risk. A forward contract turns that uncertainty into a fixed line item.

The model is a direct transplant from quantitative finance, applied to the newest raw material of the digital age.

The Team from Citadel and Revolut

The founders bring complementary backgrounds that map neatly onto the problem. Roman Yanushevskyi, an NUS Computing alumnus, worked in quantitative research at Citadel Securities, an experience that immersed him in the mechanics of electronic trading and derivative markets [NUS Computing, November 2026]. His co-founder, Ilia Bolgov, is a former product employee at Revolut, where he would have seen the challenges of scaling a fintech platform and managing complex, real-time financial flows [FundUp.ai, November 2026]. Together, they started the company in February 2026 and were accepted into Y Combinator’s Summer 2026 batch [NUS Computing, November 2026].

Founder Prior Experience Education
Roman Yanushevskyi Quantitative Research, Citadel Securities; Engineering, Lovable National University of Singapore (Computer Science & Mathematics)
Ilia Bolgov Product, Revolut Imperial College London

Their early traction with providers suggests they are speaking a language infrastructure operators understand. The reported $30 million in listed contracts, while not yet realized revenue, indicates a willingness from the supply side to test a new channel [Touchmark, November 2026].

Funding and the YC Stamp

Touchmark closed a $3 million seed round in November 2026, bringing its total disclosed funding to $3.5 million after an initial $500,000 from Y Combinator [Touchmark, November 2026][FundUp.ai, November 2026]. The seed was led by Y Combinator and included Inception Fund, Liquid 2 Ventures, Kima Ventures, and a list of angels like Datadog CEO Olivier Pomel [Touchmark, November 2026]. The investor mix points to confidence in both the infrastructure thesis and the team’s specific cross-disciplinary approach.

Accelerator (2026) | 0.5 | M USD
Seed (Nov 2026) | 3.0 | M USD

The Liquidity Challenge

The most significant hurdle for any new marketplace is achieving liquidity,the point where there are enough buyers and sellers that trades happen easily and prices are stable. Touchmark’s early metrics are requests and listings, not executed volume. The platform must convince providers to list meaningful, attractively priced capacity and, simultaneously, convince application companies to shift their procurement mindset from a direct cloud vendor relationship to a financial instrument.

  • Provider trust. Will major clouds or large GPU cluster operators see this as a valuable ancillary sales channel or as a threat to their margin and customer lock-in?
  • Buyer education. Purchasing forward contracts requires forecasting future token needs, a new discipline for many engineering teams used to paying for what they use, as they use it.
  • Contract standardization. For a true market to flourish, contracts must be standardized enough to be fungible but flexible enough to cover the myriad model, provider, and SLA combinations the market demands.

The company’s answer likely lies in its narrow initial wedge: serving the long tail of AI application companies and the emerging tier of specialized inference providers for whom a liquid market offers the most obvious benefit. Success with this cohort could create a reference network that attracts larger players.

The Cultural Question

Every new marketplace answers a cultural question about what we are willing to treat as a commodity. Touchmark’s bet is that we are ready to treat intelligence,the output of the most complex software systems ever built,as a bulk good to be traded in advance. It reframes AI from a magical, on-demand service to a planned resource, like electricity or bandwidth. The product’s spreadsheet interface is not an accident; it is the argument. It suggests that the future of AI adoption will be managed not just by researchers and engineers, but by procurement officers and financial planners, hedging their bets on the price of a trillion tokens next quarter. The market, if it works, won’t just sell compute. It will sell predictability.

Sources

  1. [Touchmark, November 2026] Touchmark raises $3M to solve the AI economy | https://touchmark.ai/blog/touchmark-raises-3m-to-solve-the-ai-economy/
  2. [NUS Computing, November 2026] With US$3 million in seed funding, NUS Computing alum Roman Yanushevskyi is selling AI tokens the way farmers sell wheat | https://www.comp.nus.edu.sg/news/with-us3-million-in-seed-funding-nus-computing-alum-roman-yanushevskyi-is-selling-ai-tokens-the-way-farmers-sell-wheat/
  3. [FundUp.ai, November 2026] Touchmark $500K Seed Funding (2026) | Investors & Contacts | https://fundup.ai/recently-funded-startups/company/298674af39e5c58b424d3ee4c7aab923e8a1e79e059d0b5c6443a4b62ac624fb/touchmark
  4. [Y Combinator, retrieved 2026] Touchmark: Market for future inference capacity | https://www.ycombinator.com/companies/touchmark

Read on Startuply.vc