Site icon The Invincible India

Startup Deep Dive : Pixis — from a failed VR startup to $209 million in AI marketing backing

The Invincible India
The Invincible India Startup Deep Dive featured graphic for Pixis.

Pixis has raised $209 million from backers including SoftBank Vision Fund 2, General Atlantic and Touring Capital to sell artificial intelligence that is supposed to make advertising budgets go further. Yet in the fiscal year its India filing entity, Leapus Technologies Private Limited, is meant to show that machine off, revenue fell rather than rose — even as the company that built it once could not get a single enterprise marketer to take a meeting.

That contradiction sits at the centre of the Pixis story: a team that pivoted twice before finding product-market fit, then convinced some of the world’s largest technology investors to fund a “codeless” AI layer for marketers, only to post a revenue decline in its most recently disclosed year even as it kept turning a profit. What happened in between, and what it means for a company selling automation to an industry that is now automating itself, is the rest of this piece.

Quick facts

Company Pixis (formerly Pyxis One); India operations filed as Leapus Technologies Private Limited
Founded 2018, renamed from Pyxis One to Pixis in January 2022
Founder(s) Shubham A. Mishra (co-founder, Global CEO), Vrushali Prasade (co-founder, CTO), Harikrishna “Hari” Valiyath (co-founder, Chief Business Officer)
Businesses Codeless AI infrastructure for marketing — targeting, creative generation and ad-performance optimisation sold as enterprise SaaS
Latest FY revenue ₹199.7 crore (≈$20.8 million at $1≈₹96.0) in FY25, per the India filing entity’s accounts (Inc42)
Latest FY profit/loss Profit after tax of ₹1.6 crore in FY25, down from ₹21.5 crore in FY24 (Inc42)
Listed Private — no IPO filed or announced
Market value / last valuation Not officially disclosed since the September 2023 Series C1 round; $209 million raised in total across four rounds
Key shareholders SoftBank Vision Fund 2, General Atlantic, Touring Capital, Celesta Capital, Chiratae Ventures, Premji Invest, Grupo Carso

What they do

Pixis sells a library of pre-built, self-learning AI models that plug into a marketing team’s existing ad accounts on Google, Meta and TikTok, rather than requiring the customer to hire data scientists to build such models in-house — the “codeless” pitch that gives the company its category name. The buyer is typically an enterprise or mid-market brand’s growth or performance-marketing team.

The origin

Shubham Mishra, Vrushali Prasade and Hari Valiyath met as students at BITS Pilani and built their first company together before Pixis existed at all. That company, Absentia, made a virtual-reality headset, and reportedly raised around $1.2 million in early backing before the founders concluded that consumer VR was, as one profile of Mishra later put it, a market that stayed five years from its moment no matter how many years passed (YesPress, accessed September 2026). The team then turned the same generative-AI skills toward gaming, building tools to auto-generate art and character assets — an idea that also failed to find paying customers.

The insight that stuck came from a different direction: the founders, all performance advertisers by then, kept meeting marketing teams drowning in manual, spreadsheet-driven ad optimisation just as the industry was losing the cookies and third-party data it had built campaigns on for a decade. If a machine could keep re-learning what worked across a cookieless web faster than a human team could manually test it, that machine — not another dashboard — was the product marketers would actually pay for. That bet became Pyxis One, and eventually Pixis.

The struggle years

The years before product-market fit were not gentle. Absentia’s VR pivot cost the founders their first outside capital and their first company, sometime before 2018, when hardware and consumer appetite for headsets never arrived on the timeline investors had underwritten. The follow-on attempt — AI-generated gaming assets — fared no better commercially, even though it kept the team’s machine-learning bench sharp (Inc42; pixis.ai/about, accessed September 2026).

Even after the founders pointed that machine-learning capability at marketing and bootstrapped the business to roughly $3 million in annualised revenue, they had no enterprise sales network and no warm introductions to the chief marketing officers who controlled ad budgets. According to an account of Mishra’s early sales motion, he spent two months showing up at the lobby of a bank, hoping for a chance conversation with someone in its marketing function, before a CMO finally asked who he was and gave him fifteen minutes (YesPress, accessed September 2026). Even the company’s current name is a scar from that period of confusion rather than a plan: the product was originally called Pyxis, but so many prospects and journalists misspelled it “Pixis” that the founders renamed the company to match, formally dropping “Pyxis One” for “Pixis” in January 2022 (TechCrunch, 18 January 2022).

The turning point

That fifteen-minute lobby conversation is the turning point at the founder-story level: it turned into a free pilot, and the pilot turned into roughly $1.2 million in annual revenue from a single account, the proof point that let the team pitch enterprise logos instead of small and mid-sized businesses (YesPress, accessed September 2026). The turning point at the company level came later and is easier to verify from disclosed numbers. Pixis closed a $17 million Series B in September 2021, and only four months after that, in January 2022, it closed a $100 million Series C led by SoftBank Vision Fund 2, with General Atlantic joining as a new investor alongside returning backers Celesta Capital, Premji Invest, Chiratae Ventures and Crimsonox Ventures (General Atlantic, January 2022; TechCrunch, 18 January 2022).

The before-and-after is stark. Going into that round, Pixis had raised about $24 million in total, run roughly 40 proprietary AI models, and served just over 100 enterprise customers. Coming out of it, total funding stood at $124 million, the company had committed to building out 200-plus customisable AI models, and it had the balance sheet to expand from a scrappy, founder-led sales motion into a company that global technology investors were prepared to back at scale (General Atlantic, January 2022).

The money behind it

Each lead investor changed something specific about the company. SoftBank Vision Fund 2’s $100 million cheque, more than four times the size of every prior round combined, gave Pixis the balance sheet and the brand association to sell into large enterprises rather than SMBs. General Atlantic, which entered alongside SoftBank in 2022 and then returned in the 2023 round, gave the company a growth-equity investor with staying power across two rounds rather than a one-time backer. Touring Capital, an investor focused on enterprise AI, led the 2023 round specifically to fund research and development, publisher and ad-network infrastructure, and expansion of the sales team into North America and Latin America, alongside potential acquisitions (TechCrunch, 14 September 2023). No valuation has been publicly confirmed for any round; the company has not disclosed one since the 2023 raise.

How it makes money

Pixis runs an enterprise SaaS model: brands pay a subscription to access its library of AI models rather than paying Pixis a share of the ad spend those models manage, at least based on public statements — the company has not published a take rate or usage-based pricing schedule. Money comes in as platform subscription fees, mainly from mid-market and Fortune 2000 marketing teams. Money goes out on cloud compute to train and continually retrain “self-evolving” models, on research and development to keep pace with constant changes to Google, Meta and TikTok’s ad-serving APIs, and on the sales and business-development headcount needed to sell a technical product into non-technical marketing departments — the company said in 2023 it had around 350 employees, weighted toward those two functions (TechCrunch, 14 September 2023).

The part people tend to get wrong is that “codeless” for the customer does not mean cheap for Pixis. Every policy change an ad platform makes to its bidding or targeting algorithms is a maintenance cost for a vendor promising to keep 200-plus models current on top of it — which is one plausible explanation for why profit fell so much faster than revenue between FY24 and FY25.

The numbers

The table below reflects the audited-style financial filings of Leapus Technologies Private Limited, the entity through which Pixis’s India operations are booked, as analysed by Inc42’s financial tracker. Figures are in ₹ crore and are a subset of the company’s finances, not a global consolidated statement.

Fiscal year Revenue (₹ crore) Profit / (loss) after tax (₹ crore)
FY23 161.2 (35.4)
FY24 227.0 21.5
FY25 199.7 1.6

Where the money comes from

The surprise in the split is scale of data rather than geography. In an independent analysis it published in 2025, Pixis said it had examined $996 million of Google Ads spend across 100 consumer brands between January 2024 and February 2025 (ppc.land, September 2025) — a single benchmarking exercise larger than the entire annual ad budget of most individual advertisers, and a reminder that a “codeless AI” vendor’s real asset is less the software than the aggregated spend data flowing through it.

The risks

The takeaway

The most transferable lesson from Pixis is not about artificial intelligence at all. A team with strong technical skills and two failed companies behind it — a VR headset, then AI-generated game art — still could not get an enterprise marketing budget to look at their next idea until one person, in one lobby, over two months, forced a single conversation to happen. Capital, credibility and a $100 million round from SoftBank followed years later, but only after that one relationship turned into revenue the founders could point to. The order matters: distribution earned the right to raise money; the money did not create the distribution. For any founder building something genuinely useful but unable to get past a gatekeeper, the model is not a better pitch deck — it is showing up long enough that a single senior person finally asks who you are.

Frequently asked questions

What does Pixis actually sell?

Pixis sells a subscription to a library of pre-built AI models — covering audience targeting, creative generation and ad-performance optimisation — that enterprise marketing teams plug into their existing Google, Meta and TikTok ad accounts without needing an in-house data-science team.

Who founded Pixis, and what did they build before it?

Shubham A. Mishra, Vrushali Prasade and Hari Valiyath, who met at BITS Pilani, founded Pixis. Before it, the same team built Absentia, a virtual-reality headset company, and then a generative-AI tool for gaming assets; neither found a lasting market before the team pivoted to AI for marketing.

How much funding has Pixis raised, and who are its main backers?

Pixis has raised $209 million across four rounds: a $7 million Series A (2020), a $17 million Series B (2021), a $100 million Series C led by SoftBank Vision Fund 2 (January 2022), and an $85 million Series C1 led by Touring Capital (September 2023). General Atlantic, Celesta Capital and Chiratae Ventures have backed multiple rounds.

Is Pixis profitable?

Its India filing entity, Leapus Technologies Private Limited, reported a profit after tax of ₹1.6 crore in FY25, down sharply from ₹21.5 crore in FY24, after posting a ₹35.4 crore loss in FY23. Separately, the company said in 2023 that it expected to reach profitability globally by the fourth quarter of that year, though that claim is company-stated rather than audited.

Why did Pyxis One become Pixis, and is the company a unicorn?

The company renamed itself from Pyxis One to Pixis in January 2022 because most people misspelled “Pyxis” with an “i”. No valuation has been publicly confirmed since its September 2023 funding round, and the company has not been reported by its investors or in its own disclosures as having reached a $1 billion valuation, so it should not currently be described as a unicorn.

Sources

Figures are as of September 2026. Currency converted at $1 ≈ ₹96.0 as of 18 September 2026 (Trading Economics).

Found an error? Write to us and we’ll correct it in the open, dated, on the piece.

Exit mobile version