New startup ideas · AI and software · AI infra and compute

startup idea

Fluxbid

An autonomous agent that forecasts, buys, and resells your AI compute.

Fluxbid is an ML agent that enterprises connect to their cloud and inference billing with read access; it forecasts workload demand, buys GPU hours and future inference capacity across providers and trading venues, routes traffic to the cheapest qualified backend, and resells surplus commitments.

AI agent as a service
Enterprise
$10-100B market
Network effects, winner takes most
Global from day one

4/5

venture judge

similar startups, last 2 years (67 all-time)

99%

of 4 nearest real companies still alive

yes

8 matching federal grants and programs

Direction supported by government programs and grants

Test it before you build it

$900 · 4 weeks · 25 prospects

For $900 and 4 weeks, prove that AI companies spending $50k+ a month on inference will hand over billing data, that a 15%+ price spread actually exists in their bills, and that 3 of them will pay a $1,000 deposit against a savings-share contract.

Riskiest assumption · A head of infrastructure spending $50k+ a month on inference believes double-digit savings exist that they cannot capture in-house, and will grant a third party read access to billing and pay 25% of verified savings rather than negotiate committed-use discounts directly with their providers.

1Focus group: who and where

The head of infrastructure, platform engineering lead or CTO at a 20-200 person AI product company with a $50k+ monthly inference and GPU bill spread across two or more providers, who just got asked by finance to explain why the model bill doubled this quarter.

where to find 25 · MLOps Community Slack (20k+ practitioners, post in #discussions and DM infra leads); a list of Series A-B AI product companies pulled from the YC startup directory and Crunchbase, filtered to companies shipping LLM features with 20+ engineers; speaker and sponsor rosters from AI Engineer World's Fair and Ray Summit, where the people on stage run exactly this scale of inference; plus warm intros from your own investors into their AI portfolio companies.

2Sell first, build later

A read-only savings audit delivered in 10 business days from 90 days of billing exports, then a savings-share agreement: their inference and GPU spend gets repriced and rebooked across providers and venues, live within 45 days of signing, no workflow change on their side.

the ask · 25% of verified monthly savings - on a $50k bill saving 15%, about $1,875 a month - with a $1,000 refundable deposit at signature.

a real yes · A signed savings-share agreement with the $1,000 deposit paid is a yes. Shared billing data, praise for the audit, and 'we'll revisit next quarter' are not.

3Small experiments

The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.

1. Billing access ask
$250 · 10 days
Pull 40 names of infra and platform leads from the YC directory and Crunchbase, DM another batch through MLOps Community Slack, and ask each for a 20-minute call plus 90 days of billing exports (OpenAI, Anthropic, AWS, GPU clouds) under a mutual NDA. The founder runs every call with a 5-question script covering monthly spend, provider mix, and who owns the bill.
keep going if · 10 of 25 contacted book the call and 6 of 25 hand over billing exports under NDA
2. Hand-built savings audit
$150 · 10 days
For the first 5 datasets, rebuild the last 90 days of spend in a spreadsheet and reprice every line against current provider price sheets, committed-use tiers, and quotes requested from Computable and Touchmark - no code, about two founder-days per company. Deliver each company a one-page audit naming the dollar savings and exactly which workloads move where.
keep going if · 3 of 5 audits show 15%+ verifiable savings on real bills; under 8% median means the spread the business monetizes is not there
3. Deposit-backed savings share
$500 · 8 days
Walk each audited company through its numbers on a 30-minute call and put a 2-page, lawyer-reviewed savings-share agreement on the table: 25% of verified savings, $1,000 refundable deposit due at signature, go-live within 45 days. No custom MSA, no redlines beyond the template.
keep going if · 3 of 5 audited companies sign and pay the $1,000 deposit

4Collect a deposit up front

Tesla took $1,000 refundable reservations for the Model 3 and $100 for the Cybertruck before building either: the deposit is the measurement, not the revenue.

$1,000

per prospect, refundable

how · A $1,000 refundable deposit invoiced with the savings-share agreement and paid by ACH or card, credited in full against the first savings-share invoice; a mid-market infra lead can approve $1,000 without procurement, and the payment separates buyers from audit tourists. The CTO or head of infrastructure signs the 2-page agreement. set up: Stripe Invoicing ↗

what it reserves · One of 5 first-cohort slots, the audit's repriced rates locked for 90 days, and go-live within 45 days of signing.

refund · Refunded in full within 5 business days on request before go-live, or if verified savings come in under 10% in the first 60 days.

target · 3 deposits from 5 audit-backed proposals (25 prospects reached) within 28 days.

Go: build it if

6 of 25 prospects share billing data, at least 3 of 5 audits show 15%+ verifiable savings, and 3 signed agreements with $1,000 deposits paid inside 4 weeks - build the forecasting agent.

Kill: stop if

Fewer than 4 of 25 share billing data, median audited savings under 8%, or 0 deposits after 5 audit-backed proposals - either the trust barrier or the price spread kills it.

5 Scripts to run itoutreach message, landing copy, deposit terms · click to open

outreach message

You're running $50k+ a month of inference across providers, and I'd bet 15-25% of it is priced wrong - wrong tier, wrong model mix, unclaimed committed-use discounts. I'm doing free savings audits for 5 AI companies this month: you share 90 days of billing exports under NDA, I hand back a line-by-line repricing in 10 business days. If I find nothing, you keep the report. If I find real money, we talk about capturing it. Do you have 20 minutes this week?

landing page

Your inference bill is 15% too high. We find it free. Free 10-day audit from your billing exports; then 25% of verified savings, you keep 75%. Book one of 5 September audit slots.

deposit terms

A $1,000 refundable deposit reserves one of 5 first-cohort slots and is credited in full against your first savings-share invoice. It locks your audited rates for 90 days, with go-live within 45 days of signing. Refunds are issued within 5 business days on request before go-live, or automatically if verified savings come in under 10% in your first 60 days.

Would you run this test?

One tap. The yes-share feeds the Demand pillar of this idea's score; nobody sees who answered.

Budgets are out-of-pocket estimates for a team of one to three, US market. Size the deposit to the deal, and check the terms before taking money in a regulated line.

Scorecard

One score that balances how trendy the idea is, the demand for it and its potential for 100x, with competition measured relative to every other idea in the catalog. Recent startup trends first, government priorities second.

Idea Score, 0-100 · raw 44.1 x 1.61

Crowded

competition: more crowded than 86% of ideas · headwind x0.57

+4.3

government priorities, secondary (67 matching grants)

Trend

Is the wave forming now? 2025-26 entrants vs 2023-24, rounds since 2025, the sector's live-batch direction, the 2026 trend analyst.

Entrants 2025-26 vs 2023-24 (similar companies)90
Rounds announced 2025+ in the sector82
Sector direction (live batch)100
2026 trend analyst50

Demand

Does anyone want it? YC's current RFS, companies already paid for something similar, the operator judge, founders' yes-rate in decks, readers who would run the test.

YC asks for it (current RFS: idea / sector)30
Someone already pays (similar companies, recent / all-time)100
Operator judge: real pain75

100x potential

Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.

Venture judge75
Market size axis67
Moat axis100
Neighbours still alive32
Technologist judge25

Score = 100 x cbrt(Trend x Demand x 100x) x (1 - 0.5 x crowding) + government bonus (max 5), calibrated so the 95th-percentile idea scores 90 (order never changes). A geometric mean: a weak pillar cannot be papered over. Percentiles are among the 272 ideas in the catalog; the terms matched were autonomous, forecasts, buys, resells, compute, enterprises, connect, cloud.

The idea in full

What: Fluxbid is an ML agent that enterprises connect to their cloud and inference billing with read access; it forecasts workload demand, buys GPU hours and future inference capacity across providers and trading venues, routes traffic to the cheapest qualified backend, and resells surplus commitments. Onboarding is self-serve from billing APIs, and pricing is a share of verified savings, so margins are pure software. The deep work is demand-forecasting and price-prediction models trained on cross-customer telemetry.
Why now: Conifer (S26) claims 80%+ token-spend reduction from least-cost routing and Understudy Labs claims 80% LLM-bill cuts, while Touchmark is building a market for future inference capacity and Computable offers GPU hours with instant liquidity - the venues an autonomous compute buyer needs to trade against now exist.
Wedge: first customer and entry point: Mid-size AI product companies with $50k+ monthly inference bills; a read-only savings audit from billing data converts them before the agent ever places a trade.
Path to 100x: Enterprise inference spend is a $10-100B line item growing with every agent deployment, and a savings-share on double-digit percentage cuts is a multi-billion revenue pool. Aggregation is winner-take-most: the largest demand book gets institutional pricing and the richest telemetry, which trains the best forecasting models, which attracts more demand - a data moat that compounds with every connected customer.
Ceiling: Hyperscalers automating committed-use discounts natively could flatten the spread until only the largest, most volatile workloads are worth trading.
Closest real companies, as the generator saw them: Conifer and Understudy Labs optimize routing within a single customer's traffic; Tracer combines open-source models for quality; none aggregates cross-customer demand into a trading book against venues like Touchmark and Computable.
Main risk: Major model providers move to flat committed-use enterprise pricing, collapsing the price dispersion the agent monetizes.

Five judges

Each judge scores every idea in the catalog with a named rubric; the venture judge decides whether a card is shown at all (4-5 is venture-grade).

Venture investor
4/5
Software margins on a $10-100B spend with cross-customer forecasting telemetry compounds, though the spread it monetizes can collapse to committed-use pricing.
Bootstrapper
3/5
Software margins and a read-only savings audit convert fast, but savings-share pricing evaporates if providers move to flat committed-use deals.
Operator
4/5
Companies with $50k monthly inference bills already feel it, and savings-share pricing off a read-only billing audit needs no workflow change.
Technologist
2/5
Read-only billing access plus routing across Touchmark and Computable is a thin arbitrage layer providers erase with flat committed-use pricing.
Risk
1/5
Depends on nascent venues like Touchmark and Computable existing, on provider terms permitting commitment resale, and on price dispersion hyperscalers are actively closing.
trends
3/5
Venues like Touchmark and Computable existing at all is a real new precondition for a compute-trading agent, but hyperscaler flat pricing could erase the spread fast.

Similar startups in the directory

Companies whose pitch matches most of the idea's terms (autonomous, forecasts, buys, resells, compute, enterprises, connect, cloud): 67 all-time, 19 from the last two years. Same matching as Idea Check.

Rescaleyc W12 · 2012 · AI infra and computealive
High Performance Computing Built for the Cloud
Zantplugandplay PnP 2026 · 2026 · AI infra and computesite down
Zant simplifies and accelerates the deployment of AI models on edge devices, where cloud-based execution is limited by privacy, latency, or connectivity constraints.
Sona8yc F26 · 2026 · Vertical AI agentsalive
Context layer for enterprise AI surveying employees with voice agents
Sentient OSyc F26 · 2026 · Horizontal AI assistantsalive
On-device AI that knows your entire life and does your work overnight.
Agent 37yc F26 · 2026 · Agent infrastructurealive
Persistent sandboxes for agents like hermes, openclaw, claude code
Scalpelplugandplay PnP 2026 · 2026 · Healthcare and bioalive
Scalpel provides a platform to optimize surgical procedures in clinics and hospitals.
Exosatyc S26 · 2026 · Defense and spacealive
Building a neutral, sovereign alternative to Starlink
Agency Tool Companyyc S26 · 2026 · Robotics and physical worldalive
Reliable infrastructure to build, deploy and manage real-world robots.
Marbleyc S26 · 2026 · Vertical AI agentsalive
The autonomous back-of-house for restaurants.
Tolmoyc X26 · 2026 · Security and compliancealive
Tolmo is the AI security platform that secures every layer of cloud…
Klaimeeyc X26 · 2026 · Fintechalive
Liability insurance for AI Agents. You deploy agents, we cover you.
Reason Machinesyc X26 · 2026 · Vertical AI agentsalive
Reason Machines builds Reason Agent, the autonomous software engineer…

Run this as an Idea Check →

The generator's reference companies

Real companies the model named as closest when it wrote the card, with their fate. A check mark is a company the radar could verify in its directory.

Public money in this direction

US federal grants, SBIR/STTR awards and open opportunities from the radar's public-money feed, matched to the idea's terms; the sector totals give the context.

grants and programs matching the idea

243

startup-relevant grants in AI infra and compute

$339M

awarded in the sector, tracked

opportunities open now in the sector

I-Corps: Translation Potential of Cross-Platform Wearable Interface for Spatial and Ambient Computingawardhigh relevance
National Science Foundation · I-Corps · $50K · posted 2026-08-17
I-Corps: Translation Potential of an Artificial Intelligence (AI)-Enabled Multi-Sensor Vest for Real-Time Safety and Situational Awareness in High-Risk Operationsawardhigh relevance
National Science Foundation · I-Corps · $50K · posted 2026-07-24
SBIR Phase I: AI-Driven Cloud Application Programming Interface for Quantum-Accurate Materials Simulationawardhigh relevance
National Science Foundation · SBIR Phase I · $305K · posted 2026-08-11
VINES: Track 1: Enabling Real-Time Agricultural Robotics with AI-Native Cellular Access Edgeawardhigh relevance
National Science Foundation · Use-Inspired NextG · $1M · posted 2026-07-30
Collaborative Research: III: Graph-based Vector Databases, from Theory to Practiceawardhigh relevance
National Science Foundation · Info Integration & Informatics · $200K · posted 2026-07-13
Collaborative Research: III: Graph-based Vector Databases, from Theory to Practiceawardhigh relevance
National Science Foundation · Info Integration & Informatics · $400K · posted 2026-07-13
Collaborative Research: III: Graph-based Vector Databases, from Theory to Practiceawardhigh relevance
National Science Foundation · Info Integration & Informatics · $200K · posted 2026-07-13
AIS Comprehensive Radon Intelligence Platform: A Connected Ecosystem for Radon Awareness, Decision Support, and Mitigation.awardhigh relevance
NIH / NIEHS · SBIR phase I · $259K · posted 2026-07-01

All public money by sector →

Market signal

What the radar sees in AI infra and compute: new companies by cohort year, the forming YC batch, and outcomes since the February snapshot.

AI infra and compute · 11 → 34 → 59 → 58 → 72 new companies 2022 → 2026 · 95% aliveYC F26 live: 6 in this cluster, 5% of the batch (was 8% in S26)Since February, of 116 YC companies here: 3 acquired, 3 shut down, 26 rewrote their pitch

AI infra and compute: companies, trend and grants →

Design attributes

The card is one cell of a designed set: every axis below was chosen before the text was written, and the text had to realize it.

Buyer: Enterprise
Business model: AI agent as a service
Path to 100x: Network effects, winner takes most
Market size: $10-100B market
Capital intensity: Capital-light (software margins)
Speed to revenue: Revenue in 1-3 years
Technical depth: Deep tech: ML, hardware, bio
Go-to-market: Self-serve
Moat: Data moat
Geography: Global from day one
Regulation: Unregulated
Vibe: Hot space