Opportunities › GPU cloud infrastructure pricing and access › Proposed app
Tallyrack
Rent H100 and H200 GPUs by the hour, with a spend cap on every run
- Score
- 32
- confidence 46
- Evidence
- 4
- items on the page
- Rivals
- 1
- named in sources
- First version
- 45 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2010
- Mature market · AI estimate
ScreensConcept
Desktop app: ML engineers pick GPUs, launch experiments, and monitor real-time costs during development sessions at a desk or lab. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Rate board
Sample rates$2.04/hr
About this appAI proposal
Tallyrack is a console for renting high-end GPUs by the hour. ML engineers and small teams pick a GPU, launch a container or notebook, and watch the cost of the run climb in real time. Every run has a spend cap, so a forgotten job cannot turn into a surprise bill.
It is built for startups, researchers and developers who need H100 or H200 time for training and inference but have no budget for committed contracts. Capacity comes from supply partners, and the console compares their hourly rates in one list.
What is different: price and cost control come first. You see the hourly rate, the cap, the idle timer and the running total on one screen, from phone or desktop, and you can stop a run from anywhere.
- Tagline
- Every GPU hour, counted and capped.
- Voice
- PlainPreciseFrugal
- Why this name
- A tally counts hours and cost; a rack is where GPUs sit. Two plain words that say the product keeps count of compute.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Live rate board
Compare hourly rates for H100 and H200 across supply partners in one list, with availability shown next to each price.
No member yet
Hard spend caps
Set a dollar or hour limit on each run. When it is reached, the run is checkpointed and stopped.
No member yet
Idle auto-stop
Detects no GPU activity for a time you choose and pauses the instance so you stop paying for idle hours.
No member yet
One-tap launch
Start from a container image or notebook template, with a disk volume that persists between runs.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#0F5C5ADeep Teal · primary#F6F3ECPaper · surface#1E2A2BInk · text#E2711DSignal Orange · accent#3B6EA5Slate Blue · data
Typography
Aa
Tallyrack
ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789
Every GPU hour, counted and capped.
PricingAI proposal
The material gives no verified competitor prices. The only rates are a public claim on Hacker News (H100 $2.04/hr, H200 $3/hr). GPU rates here are a starting point to test; the Team fee is a guess.
Pay as you go
Free
- No console fee
- GPU time billed per hour used
- Spend caps and idle auto-stop
- Run cost ledger
Team
$49 a month
- Everything in Pay as you go
- Project budgets and members
- Phone alerts for the team
- Export of cost reports
Would you pay Team · $49 a month?
Your answer tells PWAS which prices people accept. It changes no score.
No answers yet.
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- No loop in the product. Reaches customers through sales, content and ads. Public rate comparisons and cost-per-run write-ups can draw ML engineers, but nothing in the product's own use exposes it to new buyers.
- Do they become customers?
- Test whether engineers who find a cost write-up or rate board will fund a first run and come back. Team invites might help later, but they are not a loop to count on yet.
- First customers
- Seed-stage AI startups with two to ten engineers that fine-tune open models and pay on-demand GPU rates today, with no committed contract.
- Where to reach them
- Hacker News and Reddit ML: Post a transparent rate comparison and a cost-per-run write-up, with a link to launch a first run.
- Open-source ML Discords: Offer small trial credits to people running fine-tunes and ask for feedback on caps and alerts.
- Direct outreach: Message founders of small AI startups who post about GPU bills and offer a test run on their workload.
- First milestone
- Ten teams each spend over $200 on GPU time in a month, and at least half return for a second month.
Why people already pay for this
- VerifiedHacker News post received 63 points and 20 comments
- Public claimH100 GPUs available at $2.04 per hour; H200 GPUs at $3 per hour
- Public claimService positions itself as the cheapest GPU cloud option
- Public claimThere is demonstrated demand for GPU cloud services (63 points and 20 comments on Hacker News)
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · GPU supplyContracts or APIs with GPU hostsThe product needs real H100 and H200 capacity at a low rate; the PWA only controls and bills it.
- This app · Instance controlContainer launcher with SSH and notebook gatewayStarts, stops and checkpoints runs on partner hosts and reports GPU use for idle detection.
RisksAI inference
- CommercialThe low-price claim is unverified. Margin may vanish if host rates rise or if the supply partners sell directly to the same teams at the same price.
- PlatformDepends on supply partners for capacity. Stock-outs, changed terms or a partner leaving would break launches and the rate board.
- TechnicalReliable checkpoint on cap, idle detection and cost metering across different hosts is hard. A missed stop means the customer is billed for unwanted hours.
- RegulatoryHolding customer data and running their code on third-party hardware raises terms, abuse and data location questions to settle before launch.
- CommercialPrepaid or card-on-file GPU rental attracts fraud and crypto abuse, which can lead to chargebacks and account reviews.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.