The Week Software Started Charging for Results
SRF Capital Studio Research DeskFunding Intelligence, SRF Capital StudioIndian startups raised $203.4 Mn across 21 deals, and the week's clearest idea was that software can now be paid for finishing the job
Summary
- Indian startups raised $203.4 Mn across 21 deals, 3.4 times the $58.9 Mn of the week before. Two rounds are about 80% of it.
- The average disclosed round was $12.7 Mn. The median was $1.8 Mn. Benchmark against the median.
- Four of the five AI companies funded this week sell work done rather than software used. Ema says it charges per completed task, not per seat and not per token.
This week in numbers
| Figure | What it is |
|---|---|
| $203.4 Mn | Raised across 21 deals |
| 3.4x | The week before, at $58.9 Mn |
| $12.7 Mn | Average disclosed round |
| $1.8 Mn | Median disclosed round |
What is the customer actually paying for?
Every business sells one of two things. It sells a tool, or it sells a result. A hammer is a tool. A carpenter who fixes your door is a result. You pay for the hammer whether or not the door gets fixed. You pay the carpenter when it does.
For four decades, software sold tools: first a licence bought once, then a subscription priced per user, the "seat" model behind most SaaS companies. The customer still did the work.
Services firms sold results. India built one of the world's largest industries this way, with IT services and outsourcing companies billing for people, hours and tickets closed.
AI agents are erasing that line. An agent is software that does not wait for a person to click through every step. It plans a task, works across other applications, checks its own output, and finishes the job, returning to a human only for judgment or approval. Once software can finish the job, it can be paid for finishing the job.
Figure 1 / How software gets paid, and who carries the risk
- Licence: pay once for the toolThe customer. It works or it does not, the money is spent.
- Seat subscription: pay per user per monthShared. The vendor must keep you renewing, you must get value from the seats.
- Outcome pricing: pay per task completedThe vendor. No completed task, no payment.
That idea is the thread through this week. The clearest example is Ema's $77 Mn . Ema says it charges neither per seat nor per AI token, but per completed task and business outcome. Four of the five AI companies funded this week sell some version of "work done" rather than "software used." Beyond AI, in motorcycles, satellites and medicine delivery, the same three questions surfaced: what is the outcome, who verifies it, and what does it cost to deliver?
Once software can finish the job, it can be paid for finishing the job.
A big number with a narrow base
Indian startups raised $203.4 Mn across 21 deals between September 21 and 25, 3.4 times the $58.9 Mn raised across 14 deals the week before.
The headline hides a narrow base. Two rounds, Ultraviolette at $85 Mn and Ema at $77 Mn, add up to $162 Mn, or about 80% of the week. The other 19 deals shared roughly $41 Mn. The average disclosed round was $12.7 Mn. The median was $1.8 Mn.
Headlines describe the average. Your own fundraise will look much more like the median.
Figure 2 / The fourteen largest of sixteen disclosed rounds
| Startup | Round size (USD Mn) |
|---|---|
| Ultraviolette | |
| Ema | |
| Rivet | |
| Dextr AI | |
| GalaxEye | |
| Rio Health | |
| Sol Foundry | |
| Definedge | |
| ONYA | |
| betterhood | |
| ByteAsk | |
| 1312 Interactive | |
| Protein Pantry | |
| Demoverse |
Figure 3 / Weekly funding, eight weeks to 25 September
Total funding (USD Mn)
- 247.43 Aug
- 139.510 Aug
- 233.217 Aug
- 210.324 Aug
- 176.531 Aug
- 321.97 Sept
- 58.914 Sept
- 203.421 Sept
By sector, AI led with $89.3 Mn across five startups, and Ema alone accounted for 86% of that. Cleantech followed with $85 Mn from a single deal. Ecommerce had the most deals, six, but raised only $3.8 Mn between them. Split by business model, consumer companies raised $110.2 Mn and business-facing companies raised $92.2 Mn. Remove Ultraviolette, and consumer funding falls to about $25 Mn.
Figure 4 / Where the week's capital went, by sector
- AI89.3 · 44%
- Clean tech85 · 42%
- Media and entertainment10.5 · 5%
- Spacetech6.7 · 3%
- Everything else11.9 · 6%
Figure 5 / The week, by announcement date
- No deals
- Sol Foundry
- Definedge
- Protein Pantry
- Demoverse
- Ultraviolette
- betterhood
- Indian Walker
- TimBuckDo
- Space Philic
- Ema
- Rio Health
- ONYA
- ByteAsk
- Nuvah
- YNOS
- MyRx
- Unniyarcha
- Rivet
- Dextr AI
- GalaxEye
- 1312 Interactive
The anchor deal: Ema, $77 Mn Series B
The deal. Ema announced a $77 Mn Series B on September 23, led by Bengaluru-based Creaegis, with existing investors Accel, S32 and Prosus all increasing their stakes. The round was entirely primary , with no debt and no of founder or early-investor shares. Total funding now stands at $140 Mn.
The valuation. Undisclosed. Ema says its valuation has more than quadrupled since its previous round. One trade outlet reported that earlier talks centred on roughly $800 Mn. We have not been able to verify that figure anywhere else, and we would not plan against it.
What it does. Ema was founded in 2023 by Surojit Chatterjee, formerly of Google and Coinbase, and Souvik Sen, formerly of Okta. Its "AI Employees" coordinate multiple agents to complete multi-step work in HR, IT and finance inside the software a company already uses. Named customers include Wipro, Hitachi, ADP and PwC, and Ema says it handles 2.9 million employee queries a year at Wipro alone.
Figure 6 / Ema at Series B
- $77 MnRaised, all primary equity
- $140 MnTotal funding to date
- 180%Net dollar retention, company-reported
- 80%Gross margin, company-reported
The numbers, explained from the ground up
- Revenue growth. Ema reports 50x growth over two years. Company-reported and not independently verified.
- Bookings above $150 Mn. Bookings are the full value of signed contracts, including future years of multi-year deals. They are not this year's revenue. A three-year contract worth $3 Mn counts as $3 Mn in bookings on day one.
- Net dollar retention of about 180%. Take every customer who was paying Ema a year ago. Today that same group pays 1.8 times as much. Customers only expand what works.
- Gross margin close to 80%. For every $100 of revenue, about $80 remains after the direct cost of delivering the service.
Why outcome pricing and an 80% margin belong in the same sentence. A services firm that resolves 1,000 HR tickets needs roughly ten times the people to resolve 10,000. Its margin is capped by salaries. Software that resolves tickets needs more computing power, not more people, as it scales. If Ema can hold its margin while being paid per resolved task, it earns services-style revenue with software-style economics. In our view, that is what the investors are paying for.
Our read on the risk. Outcome pricing moves risk from the customer to the vendor. If a task is not completed, the vendor is not paid. So the hardest work sits in the contract: defining what "done" means for each workflow. If that definition turns bespoke for every client, a software company slowly becomes a services company.
A related signal from the week: CodeKarma is in talks to raise $5 Mn to $6 Mn. It builds tools to manage the growing complexity of AI-generated code, and the round is expected to be led by existing investors Prosus and Accel. The same investors are backing both the agents that write work and the tools that keep that work maintainable.
Headlines describe the average. Your own fundraise will look much more like the median.
Four small bets, one shared question
The four smaller AI rounds this week are very different products. Look closely and they answer the same question: before the work reaches the customer, who checks it?
Dextr AI, $6.7 Mn Seed. Led by Elevation Capital with Foundation Capital participating, this is Dextr's first institutional round. Founded in July 2025 by Sajid Shariff and Scott Arnold, Dextr builds hotel agents for guest messaging, reservation calls and staff coordination. The company reports use at properties under brands including Hilton, Wyndham, Best Western and IHG, handling more than a million interactions a month. The outcome link is direct.
Shariff says Daisy, its agent, handles $100,000 to $300,000 of monthly bookings at one large property, a founder claim we have not seen verified independently. Hotels are not buying a chatbot. They are buying answered calls that turn into room nights. One caution: the plan includes hiring more customer-facing deployment engineers.
Sol Foundry, $4 Mn. The backers are General Catalyst, Nexus Venture Partners, DeVC, Peercheque and CRED founder Kunal Shah. Sol was founded in 2025 by three former CRED executives, Anish Karan, Prateek Srivastava and Ranjith Nair. It reads a user's Gmail for commitments such as "I'll send the deck," then starts that work on its own, drafting documents, decks and replies. Nothing is sent without the user's approval.
One detail in the use of funds deserves attention: Sol lists LLM processing costs as a spending line. For AI companies, every task completed carries a real computing bill. Inference is the new cost of goods sold, and founders should model it that way from day one.
ByteAsk, $1 Mn Pre-Seed. Backed by Y Combinator and Entrepreneur First, ByteAsk builds coding agents for C and C++, the languages behind trading systems, defence, automotive and semiconductors. Its defining feature is verification. Before a developer sees a suggested change, the agent compiles it with the project's own settings, runs the tests and passes it through memory and threading checks. The deal terms are worth understanding.
YC's published standard deal is $500,000, in two parts. The first is $125,000 for 7% on a SAFE, which implies about a $1.8 Mn post-money valuation on that piece. The second is $375,000 on an uncapped SAFE that converts at the next round's price. EF's India programme invests $125,000 for 8%. On those published terms about $375,000 of the round came from angels, many working at quantitative trading firms, who are also likely users. That split is derived from the two programmes' standard terms rather than disclosed by the company.
Demoverse, $600K Pre-Seed. Led by Lumikai through its Pixels programme, Demoverse lets consumer brands put AI-generated product designs in front of shoppers, who vote on which ones get made. Creators behind winning designs earn royalties. Here the verifier is the customer.
The pattern. Ema's agents check their own work and route approvals. ByteAsk checks against a compiler. Sol waits for the user's sign-off. Demoverse asks the market. In our view, the durable advantage in agentic AI will not come from the underlying model, which every competitor can rent. It will come from the verification layer: the system that proves the work is correct before anyone pays for it.
Figure 7 / Who checks the work?
Outcomes you can touch: motorcycles, satellites and medicine
The outcome lens is even older in the physical world. A factory is paid when a vehicle is delivered. A satellite is valuable when it works in orbit. A pharmacy is paid when the medicine arrives. This week's hardware and commerce deals show what capital is being asked to convert.
Ultraviolette, $85 Mn Series E. Led by deep-tech fund Yali Capital and TDK Ventures, with Intel CEO Lip-Bu Tan investing personally and joining as an adviser, alongside existing investors. The round is worth about ₹807 Cr. The valuation was not disclosed; the last publicly reported figure was $300 Mn in August 2022, which is now four years old.
Start with the fundamentals. In FY25, Ultraviolette's operating revenue rose to ₹32.3 Cr from ₹15 Cr, while its loss widened to ₹116 Cr from ₹61.6 Cr. The fresh round is about 25 times last year's revenue, and last year's loss was about 3.6 times revenue. This is a bet on scale, not on current earnings.
Figure 8 / Ultraviolette: revenue against loss
| Year | Operating revenue (₹ Cr) | Net loss (₹ Cr) |
|---|---|---|
| FY24 | ||
| FY25 |
The scale story is concrete. Reported current capacity is about 50,000 to 60,000 vehicles a year, with a new plant targeting 250,000 in its first phase and up to 500,000 at full build. Those capacity figures each come from a single outlet, and the 500,000 is the company's own. The company already sells in 20 European countries, targets the US in 2027, expects to be margin-positive by the end of this fiscal year and -positive in FY28. The test from here is conversion: installed capacity has to become delivered, serviced, repeat-purchased vehicles.
GalaxEye, ₹63.84 Cr from the RDI Fund. Inc42 counts this at about $6.7 Mn. It is not a venture round. The Technology Development Board, under the Department of Science and Technology, will provide the money through optionally convertible debentures. An OCD is a loan that the holder can choose to convert into shares later on agreed terms. Neither side has to fix a valuation today.
Figure 9 / GalaxEye: how the project is funded
- RDI support, via optionally convertible debentures63.84 · 26%
- Balance of approved project cost183.85 · 74%
The project has an approved cost of ₹247.69 Cr, so the RDI support covers about 26%. It will fund an indigenous OptoSAR satellite, combining radar with optical imaging so it can see through cloud and at night, at resolution finer than 0.5 metres. The goal is to move from TRL 6 to TRL 9. In plain terms, TRL 6 means a prototype has worked in a realistic test; TRL 9 means the system is proven in live operation. That gap is where most deep-tech companies run out of money, and it is exactly the gap this fund is designed to bridge.
GalaxEye has raised nearly $27 Mn to date. Under the ₹1 lakh Cr RDI Fund, one outlet reports that TDB has approved 22 projects with ₹2,192 Cr of support so far.
Rio Health, $4.5 Mn Pre-Series A. Led by Version One Ventures, with existing investors Xeed Ventures, Good Capital and Amplify, in an all-equity round of ₹43.08 Cr. Version One's share was ₹21.30 Cr. One outlet, citing regulatory filings, puts the post-money valuation at about ₹170 Cr; we have not seen that confirmed elsewhere. Total funding is now ₹62.24 Cr.
Rio delivers medicines across Delhi NCR in 20 to 30 minutes. It does more than 30,000 orders a month from three dark stores with roughly 80% repeat customers and carries around 20,000 products, with pharmacist checks and no substitutions allowed. Orders arrive on WhatsApp as text, voice or prescription photos, parsed by AI. The plan is to go from 3 to 15+ stores and reach an annual run rate above ₹150 Cr within 9 to 12 months.
Figure 10 / Rio Health: from today to target
- 3 stores, 30,000+ orders a monthAbout 10,000 orders per store per month
- 15+ stores, within 9 to 12 monthsThe same density, five times over
- ₹150 Cr annual run rate₹12.5 Cr a month, about ₹83 lakh per store per month
Now the arithmetic. ₹150 Cr a year is ₹12.5 Cr a month. Across 15 stores that is about ₹83 lakh per store per month. If each store keeps today's density of about 10,000 orders a month, the implied average order is roughly ₹830. That order value, and whether new stores reach that density quickly, is what to track.
Protein Pantry, ₹9 Cr Seed. Led by Sharrp Ventures, with Peercheque, Consumer Collective by Atrium, Indian Silicon Valley Capital and founder-angels including Varun Alagh of Honasa. The brand makes high-protein vegetarian frozen foods. Most of the money goes into its own manufacturing unit. Controlling production means controlling the outcome: recipe, nutrition and price. Note who wrote the cheques: operators who built consumer brands themselves.
The counter-signal: Rivet bets on people
In a week about software replacing human effort, the largest seed round went the other way. Rivet raised $10.5 Mn, about ₹100.6 Cr, from Peak XV Partners, Shine Capital and Blume Ventures, alongside a nationwide US launch. Founder Taru Kapoor spent close to a decade at Tinder and other Match Group brands. Her diagnosis of the category is sharp: "Instead, we removed the people."
Rivet lets people sign up as "matchers" who score potential couples for others, including people who are not dating themselves. Pairs with high scores get introduced, and both sides decide independently whether to connect.
Two points for founders. First, the outcome lens still applies: the product is judged by matches that lead somewhere, and human judgment is the verification layer. Second, this is two Indian funds writing a large seed cheque into a US consumer product with engineering in Delhi NCR. Indian capital following Indian founders to global markets is a pattern to keep watching.
Instead, we removed the people.Taru Kapoor, Founder, Rivet
The rest of the ledger
Definedge, an investment-tech platform, raised $2.3 Mn in a from angels. ONYA raised about ₹12.5 Cr in a pre-Series A led by Divisa . D2C brands betterhood ($1.2 Mn, led by Sauce.VC) and Nuvah ($417K, from angels including Shantanu Deshpande) added to the ecommerce count. Gaming studio 1312 Interactive raised $1 Mn, and horizontal SaaS company YNOS raised $221K. Five deals, Indian Walker, MyRx, TimBuckDo, Unniyarcha and Space Philic, did not disclose amounts. Peercheque and Sauce.VC were the most active investors, with two deals each.
Around the deals: policy, funds and public markets
First, the Centre is considering an anchor investment of ₹15,000 Cr to ₹20,000 Cr in a proposed National Frontier AI and Compute Fund. The money would build compute capacity and support frontier AI startups. Read alongside the GalaxEye deal, the state is becoming a patient, structured investor in capital-heavy technology.
Second, Nykaa and L'Oréal's venture fund BOLD will jointly invest in Indian beauty and personal care brands. Third, the IPO pipeline stayed active. Moneyview's IPO was subscribed 6.01 times by the end of day two, closing September 28 with a tentative October 1 listing. AceVector, Snapdeal's parent, saw its ₹420 Cr IPO subscribed 23% on day one. Furlenco is preparing an FY28 IPO of ₹1,000 Cr to ₹1,200 Cr, at a target valuation near ₹7,000 Cr. Spinny has confidentially filed its draft papers with for a ₹2,500 Cr to ₹3,000 Cr issue. A working exit window supports early-stage risk-taking.
What this means
For founders. Price for the result where you can measure it, and write the definition of "done" before you write the pricing page. Put inference costs into your gross margin model now, not after your first scale-up. And if you are raising, benchmark against the $1.8 Mn median, not the $12.7 Mn average.
For enterprise leaders and MDs. The unit of purchase is changing from licences to completed tasks. The procurement question shifts from "how many users?" to "how many resolved tickets, closed invoices or booked rooms, and who audits them?" Firms that build that measurement capability early will negotiate better contracts.
For investors. Net dollar retention and gross margin at scale are the two numbers that separate an agent business from a services business wearing software clothes. In hardware and deep tech, watch the conversion ratios: capacity to deliveries, TRL 6 to TRL 9, stores opened to store-level density.
What we are watching next week
We will watch Moneyview's October 1 listing, whether CodeKarma's round closes, and whether more AI companies disclose outcome-based pricing. The question we will keep asking of every deal: what exactly is the customer paying for?
The full deal ledger, September 21 to 25, 2026
All 21 deals, largest disclosed first
| Date | Startup | Sector | Round size | Stage | Lead investor(s) |
|---|---|---|---|---|---|
| Sep 23 | Ultraviolette Automotive | Clean Tech | $85 Mn | Series E | Yali Capital, TDK Ventures |
| Sep 24 | Ema | AI | $77 Mn | Series B | Creaegis |
| Sep 25 | Rivet | Media & Entertainment | $10.5 Mn | Seed | Peak XV Partners |
| Sep 25 | Dextr AI | AI | $6.7 Mn | Seed | Elevation Capital |
| Sep 25 | GalaxEye | Spacetech | $6.7 Mn | RDI Fund (OCDs) | RDI Fund |
| Sep 24 | Rio Health | Consumer Services | $4.5 Mn | Pre-Series A | Version One Ventures |
| Sep 22 | Sol Foundry | AI | $4 Mn | Undisclosed | General Catalyst, Nexus (reported) |
| Sep 22 | Definedge | Fintech | $2.3 Mn | Pre-Series A | Angels |
| Sep 24 | ONYA | Ecommerce | $1.3 Mn | Pre-Series A | Divisa Family Office |
| Sep 23 | betterhood | Ecommerce | $1.2 Mn | Seed | Sauce.VC |
| Sep 24 | ByteAsk | AI | $1 Mn | Pre-Seed | Y Combinator, Entrepreneur First |
| Sep 25 | 1312 Interactive | Gaming | $1 Mn | Seed | Chimera VC, T-Accelerate Capital |
| Sep 22 | Protein Pantry | Ecommerce | $939K | Seed | Sharrp Ventures |
| Sep 22 | Demoverse | AI | $600K | Pre-Seed | Lumikai |
| Sep 24 | Nuvah | Ecommerce | $417K | Pre-Seed | Angels |
| Sep 24 | YNOS | Enterprise Tech | $221K | Undisclosed | Gyan Circle Ventures |
| Sep 23 | Indian Walker | Ecommerce | Undisclosed | Pre-Seed | Palette Wealth Management |
| Sep 24 | MyRx | Health Tech | Undisclosed | Seed | SteerX VC |
| Sep 23 | TimBuckDo | Consumer Services | Undisclosed | Undisclosed | Srinath Setty, Adarsh Narahari |
| Sep 24 | Unniyarcha | Ecommerce | Undisclosed | Undisclosed | Sauce.VC, Atrium.VC |
| Sep 23 | Space Philic | Spacetech | Undisclosed | Undisclosed | Undisclosed |
Deal counts and round sizes are drawn from Inc42's weekly Funding Galore and its individual deal stories. Company detail comes from Ema's own announcement, TechCrunch as carried by AgentLocker and ChainCatcher, Reworked, Mint via TradingView, EVreporter and CB Insights. Public and policy figures come from the Press Information Bureau and the Technology Development Board. The remaining rounds are from Crunchbase News, Business Wire, Indian Startup News, StartupTalky, Research Live, BW Disrupt, NexusWild, Version One Ventures, Indian Television and StartupFox. All September 2026.
This brief is for information only. It is not investment advice.
How useful was this article?
One tap. It tells us what to write more of.
About the author
SRF Capital Studio Research Desk
Funding Intelligence, SRF Capital Studio
The next one
Get what we publish next, by email.
Working notes on raising, borrowing, protecting, growing and structuring capital in India. One email a week at most, and you can leave any time.
We use your address only to send this. See our privacy policy.
We store your address to send you these emails and nothing else. See our privacy policy.
Related reading
The services behind this article
What we do about it, for founders and finance teams. Each has its own page.
Venture Capital & Private Equity
Investment readiness, fundraising strategy, diligence preparation and transaction support for founders raising institutional equity.
Market Intelligence
A continuous read on the sector: what is moving capital, regulation and demand, and what it means for this business.
