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

startup idea

Gridshift

The marketplace where AI data centers sell their flexibility to the power grid.

Gridshift turns interruptible AI training load into a tradable grid product: data centers enroll clusters, Gridshift's ML forecasts and checkpoint-aware orchestration shift or shed training jobs on minutes of notice, and the flexibility is sold into wholesale electricity markets as demand response.

Marketplace
Enterprise
$100B+ market
Creates a new category
US first

5/5

venture judge

similar startups, last 2 years (46 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

$2,000 · 5 weeks · 15 prospects

Prove that neocloud operators stuck in the ERCOT queue will pay real money to make one training cluster curtailable, for $2,000 in 5 weeks.

Riskiest assumption · Operators of AI training clusters will contractually accept curtailment of live training runs, and pay for help structuring it, in exchange for earlier interconnection - despite the industry operator's objection that labs refuse to let anything preempt a run.

1Focus group: who and where

Head of infrastructure or VP of energy at a neocloud or AI data center developer with a 10-200 MW large-load interconnection request pending in ERCOT, currently quoted 2028 or later for full energization while customers wait on capacity.

where to find 15 · ERCOT's public Large Load Interconnection Status report (a monthly directory of pending requests, with company names), the open virtual meetings of ERCOT's Large Flexible Load Task Force (where these operators already show up), the Gulf Coast Power Association Fall Conference in Austin in October, and Infrastructure Masons chapter events in Texas.

2Sell first, build later

A $15,000 curtailment readiness study for one site: a curtailment plan for one training cluster capped at 40 hours per year with 10-minute notice, the energization acceleration case to file with the utility, and a 12-month power savings model, delivered 4 weeks from kickoff.

the ask · $15,000 per pilot study, with a $3,000 deposit due at signing.

a real yes · A real yes is the $3,000 deposit paid against a signed scope with a kickoff date on it; conference enthusiasm, requests for the deck, and offers to share load data for free do not count.

3Small experiments

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

1. Curtailment acceptance interviews
$300 · 14 days
Pull 25 pending large-load requests from the ERCOT queue report, cross-reference LinkedIn for the infrastructure lead, and book 15 calls. Ask each one flat out: would you accept up to 40 hours per year of checkpoint-aware curtailment on one training cluster, with 10-minute notice, if it moved your energization date up 12 or more months? One founder runs all calls with a fixed script and a one-page savings model.
keep going if · 8 of 15 operators accept 40 hours/year of curtailment on a training cluster in exchange for 12+ months earlier energization
2. Utility counterparty check
$100 · 10 days
Call 5 people on the other side of the trade: large-load account managers at Oncor and CenterPoint, plus business development at existing ERCOT demand response aggregators such as CPower and Enel X. Confirm that aggregated data center load can clear through ERS or ancillary services today under an existing aggregator's registration, without Gridshift holding its own license first.
keep going if · 3 of 5 confirm a working registration path for aggregated curtailable data center load under an existing aggregator
3. Paid readiness study offer
$200 · 21 days
To the interviewees who accepted curtailment, pitch a $15,000 curtailment readiness study for one site: the curtailment schedule, the energization acceleration case to file with their utility, and a 12-month savings model, delivered 4 weeks from kickoff. Ask for a $3,000 deposit on a signed one-page scope. This is consulting the two founders can deliver by hand, no code.
keep going if · 2 of 10 pitched operators sign the scope and pay the $3,000 deposit
4. GCPA conference sweep
$1,400 · 3 days
One founder attends the Gulf Coast Power Association Fall Conference in Austin with the savings one-pager, targets utility large-load staff and data center developers on the floor, and books follow-up scoping calls. This tops up the pipeline if the queue-list outreach runs dry.
keep going if · 5 booked follow-up scoping calls from 10 floor conversations

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.

$3,000

per prospect, refundable

how · An invoiced deposit on a paid design-partner study: $3,000 against the $15,000 study, on a one-page scope signed by the head of infrastructure. Neoclouds are startups that pay invoices without procurement cycles, so money can genuinely move now; the study, not the marketplace, is what is sold. set up: Stripe Invoicing ↗

what it reserves · The study slot with a kickoff inside 30 days, plus a reserved place in the first ERCOT enrollment cohort once Gridshift operates under a licensed aggregator.

refund · Fully refundable any time before study kickoff, no questions asked.

target · 2 paid $3,000 deposits from 15 operator conversations within 35 days

before taking money · Do not enroll load, bid, or settle anything in ERCOT wholesale or ancillary markets before registering as a market participant or operating under an already licensed demand response aggregator; until then sell only studies and LOIs.

Go: build it if

2 paid study deposits and 8 of 15 operators accepting 40 hours/year of curtailment on training load: build the orchestration and sign an aggregator partnership.

Kill: stop if

0 deposits after 10 pitches, or fewer than 4 of 15 operators accepting any curtailment of training load: the stated risk that nothing may preempt a training run is confirmed, stop.

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

outreach message

You have a large-load request sitting in the ERCOT queue while your customers wait on capacity. We help operators enroll one training cluster as curtailable load, checkpoint-aware and capped at 40 hours a year, which supports a faster energization case with your utility and cuts your power cost. Before building anything we run a $15,000 curtailment readiness study: your queue position, your cluster, the exact hours and dollars. Can I get 20 minutes this week to walk through the numbers for your site?

landing page

Get grid power years earlier by making one cluster curtailable $15,000 readiness study: curtailment plan, energization case, and savings model for one site Book a 20-minute scoping call

deposit terms

A $3,000 deposit reserves your study slot and a place in our first ERCOT enrollment cohort. The study kicks off within 30 days, and the $12,000 balance is invoiced only after you approve the written scope. The deposit is fully refundable any time before kickoff.

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 45.7 x 1.61

Crowded

competition: more crowded than 81% of ideas · headwind x0.59

+4.1

government priorities, secondary (53 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)45
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 pain50

100x potential

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

Venture judge100
Market size axis100
Moat axis80
Neighbours still alive40
Technologist judge75

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 marketplace, centers, sell, flexibility, power, interruptible, training, load.

The idea in full

What: Gridshift turns interruptible AI training load into a tradable grid product: data centers enroll clusters, Gridshift's ML forecasts and checkpoint-aware orchestration shift or shed training jobs on minutes of notice, and the flexibility is sold into wholesale electricity markets as demand response. Data centers buy cheaper power and interconnection approval; grid operators buy a new balancing resource. It is a licensed market participant in each ISO, sold through utilities and neocloud operators rather than direct.
Why now: Power, not silicon, is the binding constraint the brief keeps surfacing: Callosum raised a $100M seed on 2026-08-20 to tackle AI compute bottlenecks, and founders are resorting to Atomarine's floating nuclear data centers at sea and Ethos Space Resources making silicon on the Moon to escape the grid queue.
Wedge: first customer and entry point: One neocloud stuck in an interconnection queue in Texas: Gridshift enrolls a single training cluster as curtailable load, which lets the utility approve a larger connection years earlier, and takes a share of the energy savings.
Path to 100x: US wholesale electricity is a $100B+ market and data centers are becoming its largest new load; a marketplace clearing even a few percent of that flow at a transaction take is a multi-billion revenue business. FERC and ISO market-participant licenses in each region take years to obtain and compound into a regulatory moat that makes Gridshift the default counterparty on both sides.
Ceiling: If ISOs standardize a simple curtailment tariff for data centers, the marketplace collapses into a compliance feature utilities offer for free.
Closest real companies, as the generator saw them: Atomarine and Ethos Space Resources route around the grid with new generation; Dipole Labs optimizes inside the cluster with optical switching; Computable trades GPU hours, not megawatts. Gridshift is the only one making the existing grid the counterparty.
Main risk: Frontier labs refuse to let anything preempt a training run, leaving only low-value inference load to trade.

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
5/5
ISO market-participant licenses plus a take on $100B+ wholesale power flow is a compounding regulatory moat on the binding constraint.
Bootstrapper
1/5
Becoming a licensed FERC and ISO market participant in each region takes years before a single megawatt clears.
Operator
3/5
Interconnection delay is a real budgeted pain, but the model requires labs to let training runs be preempted, which they refuse.
Technologist
4/5
Checkpoint-aware curtailment of live training jobs plus per-ISO market registration is hard engineering that no router or trading venue replicates.
Risk
4/5
FERC and ISO market-participant licenses in each region are genuine regulatory footing, and multiple ISOs plus utility channels avoid a single counterparty.
trends
3/5
AI load jamming interconnection queues is a genuine 2025-2026 constraint, but demand response is a decade-old product with new customers, not a new window.

Similar startups in the directory

Companies whose pitch matches most of the idea's terms (marketplace, centers, sell, flexibility, power, interruptible, training, load): 46 all-time, 13 from the last two years. Same matching as Idea Check.

Shepherd Roboticsyc F26 · 2026 · Robotics and physical worldalive
Robots for high skilled labor powering AI infrastructure
LinkLaneyc W27 · 2027 · Vertical AI agentsalive
AI-powered freight brokerage. Quote, book, and track loads in seconds.
OS3yc S26 · 2026 · Robotics and physical worldalive
Affordable, intelligent humanoid robots built to deploy at scale
Inviscid AIyc W26 · 2026 · Robotics and physical worldalive
Real-time Physics Simulations for Industrial Facilities & Data Centers
Cortex AIyc F25 · 2025 · Data for AIalive
Large-scale real-world robot & human data for embodied AI
SellRazeyc F25 · 2025 · Consumeralive
The fastest way to sell online.
Kaspixalchemist Alchemist Class 39 · 2025 · AI infra and computeunchecked
Kaspix helps OEMs ship cheaper, upgradeable edge devices faster—using swappable manufacturable analog-AI modules that cut BOM and test time without leading-edge digital chips.
NODI Energyyc X25 · 2025 · Climate and energyalive
Distributed energy infrastructure unlocking grid power to data centers
LoadexAialchemist Alchemist Class 38 · 2025 · Vertical AI agentsunchecked
AI Powered Truck Dispatcher
Sampleapp.aialchemist Alchemist Class 38 · 2025 · Developer toolsunchecked
Generate Custom POCs/Sample Apps in Minutes for your APIs and 10x Sales Cycles. In-house API validation and generation layer. In-house container tech.
HUDyc W25 · 2025 · Data for AIalive
Platform and marketplace for RL environments and AI data
shatterdomeef EF Bangalore 2025 · 2025 · Climate and energyalive
The financial infrastructure layer for modern energy and power.

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

SBIR Fast-Track: Advanced Hexagonal Boron Nitride Composite Semiconductor Packaging for Enhanced Performance, Efficiency, and Protectionawardhigh relevance
National Science Foundation · SBIR Fast-Track · $2M · posted 2026-08-24
Virtual Reality Cognitive Therapy for MCI and Alzheimer's diseaseawardhigh relevance
NIH / NIA · SBIR phase II · $889K · posted 2026-09-22
Leveraging Coded Ptychography to Deliver AI-powered Rapid Analysis of Unstained Lung Biopsy Samples at the Point of Careawardhigh relevance
NIH / NCI · SBIR phase II · $1M · posted 2026-09-01
OmicXScreen: AI/ML-Powered, DNA Methylation Profiling Test for Early HCC and Esophageal Cancer Detectionawardhigh relevance
NIH / NCI · SBIR phase I · $405K · posted 2026-09-01
STTR Phase I: Conformal sensor for radiation monitoring using organic semiconductors and conductorsawardhigh relevance
National Science Foundation · STTR Phase I · $305K · posted 2026-08-26
SBIR Phase I: Wireless Sidestep Network: Harnessing Sub-terahertz to Enable New Degrees of Freedom in AI Data Center Architecturesawardhigh relevance
National Science Foundation · SBIR Phase I · $305K · posted 2026-08-18
I-Corps: Translation Potential of an Ultra-Low-Power Microrobotic Sensing Platform for Deployable Distributed Robotic Systemsawardhigh relevance
National Science Foundation · I-Corps · $50K · posted 2026-08-17
Novel Immunoregulatory Therapeutics for Bacterial Gastritisawardhigh relevance
NIH / NIAID · SBIR phase I · $314K · posted 2026-08-13

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: Marketplace
Path to 100x: Creates a new category
Market size: $100B+ market
Capital intensity: Capital-medium (ops, field teams)
Speed to revenue: Revenue in 1-3 years
Technical depth: Deep tech: ML, hardware, bio
Go-to-market: Partners and channels
Moat: License or regulatory moat
Geography: US first
Regulation: Heavily regulated
Vibe: Boring business