• The Runway Ventures
  • Posts
  • ๐Ÿคฏ Cake Group: How a 50/50 Cap Table Killed a $1B Singapore Crypto Unicorn

๐Ÿคฏ Cake Group: How a 50/50 Cap Table Killed a $1B Singapore Crypto Unicorn

How a 50/50 cap table, the Terra/Luna crash, and a public co-founder feud collapsed Singapore's $1B DeFi unicorn from 1M users to a fire sale

New here? Join us to become the top 1% founder by learning from startup failures:

P.S. Brands like Notion, Airwallex, Deel, and Intercom sponsor The Runway Ventures to reach founders who actually convert.

๐Ÿ‘‰๐Ÿป See why they do it

Hey Founders,

Welcome to The Runway Ventures โ€” a weekly newsletter where I deep dive into failed startup stories to help you become the top 1% founder by learning from their mistakes with actionable insights.

๐Ÿ‡ธ๐Ÿ‡ฌ Thinking of setting up a company in Singapore? This free guide from Osome outlines how to build your company's foundation the right way. Youโ€™ll discover the exact registration requirements, post-launch compliance checklists, and specific frameworks required for foreign founders.

Fixing legal mistakes costs time and money. Use this roadmap to go from idea to fully incorporated without the stress.

Today's story is about a Singapore DeFi platform that hit nearly 1M users and US$1B in customer assets, then watched a co-founder war and a winding-up filing wipe out 90% of AUM in 18 months. Let's get to it! ๐Ÿš€

Today at a Glance:

  • โ˜ ๏ธ 1 Failed Startup โ†’ Cake Group

  • โš ๏ธ 2 Mistakes โ†’ A 50/50 cap table with no tiebreaker

  • ๐Ÿง  3 Lessons Learned โ†’ Build governance for the day you disagree

  • ๐Ÿ”— The Runway Insights โ†’ The new under-utilised sales trick you can steal

  • ๐Ÿ’ฐ Southeast Asia Funding Radar โ†’ Featherless.ai raises $20M (Series A) to expand serverless AI inference platform and build neutral open-source AI infrastructure for developers globally

โ˜ ๏ธ 1 Failed Startup: Cake Group

Cake Group co-founders โ€” Dr. Julian Hosp (CEO) and U-Zyn Chua (CTO)

๐Ÿš€ The Rise of Cake Group

๐Ÿ‡ธ๐Ÿ‡ฌ Founded by Dr. Julian Hosp (CEO) and U-Zyn Chua (CTO) in 2019, Cake Group was a Singapore-based crypto yield platform that made decentralised finance accessible to everyday retail investors โ€” letting them earn passive income on idle digital assets without managing private keys or smart contracts.

๐Ÿ•บ๐Ÿป๐Ÿ•บ๐Ÿป Foundersโ€™ Story

Cake started when Julian Hosp left TenX in early 2019. His last crypto venture had raised ~$80M in a 2017 ICO, then cratered when Visa cut off its card issuer. He wanted to build something that distanced itself from ICO hype and actually generated cashflow.

Around the same time, U-Zyn Chua had been quietly building. He'd just been named Singapore's first Smart Nation Fellow, served as lead architect on the Bahamas' Sand Dollar (the world's first retail CBDC), and had a working crypto cashflow prototype sitting on his laptop.

๐Ÿค๐Ÿป Chua reached out. They teamed up.

๐Ÿ”“ The bet was simple โ†’ wrap DeFi in a clean, custodial interface, and millions of retail investors could finally earn yield on their crypto.

  • The Problem โ€” ๐Ÿคฏ DeFi was powerful but completely inaccessible for normal people.

    • Users had to manage private keys, smart contracts, and decentralised exchanges directly.

    • Existing CeFi competitors like Celsius and BlockFi were opaque black boxes โ€” using customer deposits for high-risk proprietary trading.

    • Anyone wanting passive yield on idle crypto had no simple, trusted option.

  • The Solution โ€” ๐Ÿฐ Cake DeFi was a centralised platform that wrapped DeFi protocols in a clean, custodial interface.

    • Users deposited crypto, Cake deployed it into transparent on-chain yield strategies (lending, staking, liquidity mining).

    • Published regular transparency reports and a real-time Proof of Reserves dashboard.

    • Even built its own DeFiChain layer-1 blockchain to make yield generation verifiable.

๐Ÿ‡ธ๐Ÿ‡ฌ๐Ÿฐ In short, Cake DeFi made earning yield on crypto feel as easy as opening a bank account.

๐Ÿ‚๐Ÿ“ˆ After raising just $2.4 million in seed funding in June 2019 (most of it from the founders themselves), Cake rode the 2020-2021 crypto bull run like few others.

Then 2021 happened.

๐Ÿค‘ Revenue jumped 1,800% YoY to $631 million. Operating income hit $203 million. They paid out $230 million in rewards to users โ€” that year alone. All without taking a single dollar of VC money beyond seed.

By Q1 2022, Cake was so confident it launched Cake DeFi Ventures โ€” a $100 million corporate venture arm to invest in Web3, gaming, and NFTs. A 3-year-old startup, deploying $100M into other startups.

๐Ÿคฏ A few weeks later, they signed an LOI for a SPAC merger valuing the company at $1.5 billion.

๐Ÿ”๏ธ At its peak, Cake DeFi:

  • had nearly 1 million registered users

  • managed $1 billion+ in customer assets

  • did $631 million in 2021 revenue (1,800% YoY)

  • ran with 180 employees across Singapore and Kuala Lumpur

  • had a $1.5 billion SPAC valuation on the table

  • paid out $375 million in cumulative rewards by Q2 2022

That was the dream โ€” a profitable, founder-controlled crypto fintech, IPO-ready, eyeing a billion-dollar exit.

๐Ÿ“‰ The Fall of Cake Group

๐Ÿ’” And then the partnership cracked.

The crypto winter started the bleeding. Terra/Luna collapsed in May 2022, the SPAC fell apart, revenue halved. Manageable problems for a company sitting on $169 million in assets.

But by late 2023, the cracks weren't financial anymore. They were personal. And they ended up in the Singapore High Court.

๐Ÿ“Œ Hereโ€™s what happened to Cake Group:

โ

While it is regrettable that such internal disagreements have been brought into the public domain, we are working with all stakeholders to resolve this dispute expeditiously.

โ€” shared by Julian Hosp in hi statement

๐Ÿฐ๐Ÿฐ๐Ÿฐ Have your Cake and yield it too

  • Jan 2019 โ€” ๐Ÿšช Julian Hosp stepped down as President of TenX after Visa cut off its card issuer and the ICO model fell apart.

  • 7 Jun 2019 โ€” ๐ŸŽ‚ Cake Group was officially incorporated in Singapore by Hosp (CEO) and U-Zyn Chua (CTO), with $2.4 million in seed funding from the founders, John Rost, and Howard Fineman.

  • May 2020 โ€” โ›“๏ธ The team launched DeFiChain (DFI), a Bitcoin-linked layer-1 blockchain to power transparent on-chain yield generation.

  • 2021 โ€” ๐Ÿ”ฅ Cake rode the crypto bull market โ€” revenue jumped 1,800% YoY to $631 million, with $203 million in operating income and 10x user growth.

  • 9 Mar 2022 โ€” Cake launched Cake DeFi Ventures, a $100 million corporate VC arm investing in Web3, gaming, eSports, NFTs, and fintech.

  • Apr 2022 โ€” ๐Ÿš€ Cake signed an LOI for a SPAC merger valuing the company at $1.5 billion and launched its "Borrow" feature on the same trajectory of bull-market expansion.

๐Ÿฅถ Then the music stopped...

  • May 2022 โ€” โ„๏ธ Terra/Luna collapsed. Crypto entered winter. Cake walked away from the $1.5 billion SPAC, citing the market crash and unclear regulations.

  • Q3 2022 โ€” ๐Ÿ“‰ Cake posted its first operationally negative quarter since early 2020. Crypto prices were down ~70%, AUM had shrunk, but layoffs hadn't started.

  • 15 Feb 2023 โ€” Hosp published the 2022 transparency report titled "One of the Last Ones Standing".

    • 2022 revenue โ†’ $266 million (down >50% from 2021).

    • Profit โ†’ $23.5 million (down 5x).

  • 14 Nov 2023 โ€” โœ‚๏ธ Hosp announced a 30% workforce reduction โ€” 52 employees laid off, headcount cut from ~177 to 125. CTO Chua publicly opposed the cuts, saying he'd been kept out of the loop and that the layoffs had "no clear intention".

  • 1 Dec 2023 โ€” โš–๏ธ Chua filed a winding-up application in the Singapore High Court โ€” not because Cake was insolvent, but because the shareholder relationship had completely broken down.

๐Ÿคผ Hosp vs Chua

  • Feb 2024 โ€” Both founders made eleventh-hour buyout offers to each other at the same valuation. Both were declined. A previously agreed S$32 million buyout from late 2023 had already collapsed.

  • 25 Apr 2024 โ€” ๐Ÿง‘โ€โš–๏ธ Singapore High Court dismissed Chua's application. Judicial Commissioner Alex Wong said Cake was "a thriving enterprise of value" and that Chua "had gone too far".

  • 27 May 2024 โ€” Chua sold his remaining shares to Hosp at a price he later called "much lower than the original deal."

    • Finally, 6-month legal battle ended.

  • 11 Dec 2024 โ€” ๐Ÿค GSTechnologies (LSE: GST) announced a binding agreement to acquire Cake Pte Ltd and Cake DeFi UAB for an undisclosed cash sum. By then, Cake had ~700,000 registered users, but only 50,000 active, holding ~$80 million in assets โ€” down from $1B+ at peak.

  • 2 Jan 2025 โ€” ๐Ÿ“ฆ Acquisition completed. Hosp stayed for a 3-month handover, then stepped down to "focus on his family and children".

  • 5 Feb 2025 โ€” ๐Ÿ”š Bake announced it'd shut down all Singapore operations effective 1 May 2025, citing the discontinuation of Cake Pte Ltd's MAS license exemption.

๐Ÿ“‰ By the time GSTechnologies absorbed what was left, AUM had collapsed over 90% โ€” from $1 billion at peak to $80 million.

The court called Cake "a thriving enterprise of value" in April 2024. But by then, the brand had already been quietly bled out by months of public uncertainty.

๐Ÿฉธ The product worked. The market eventually recovered. But trust between the two co-founders never did โ€” and once that breaks, no court ruling can put a startup back together.

Want to learn more about Cake Groupโ€™s downfall?

โš ๏ธ 2 Mistakes

Mistake 1: A 50/50 cap table with no tiebreaker

โ

While it is regrettable that such internal disagreements have been brought into the public domain, we are working with all stakeholders to resolve this dispute expeditiously.

โ€” shared by Julian Hosp in this statement

Cake was incorporated in June 2019 with Hosp and Chua as clean equals โ€” 50/50 voting, no institutional VC, no independent board members, no class-A/class-B share structure with deadlock provisions.

๐Ÿค๐Ÿป The logic was sound on paper. They'd raised just $2.4M in seed (mostly from themselves, plus angels Rost and Fineman), kept the cap table founder-controlled, and rode the bull market all the way to $631M revenue without a single VC dilution.

In a bull market, this looked like genius. They kept all the upside. No board to answer to. No outside agenda.

But a 50/50 split without a tiebreaker is a corporate time bomb.

๐Ÿšจ When trust collapsed in September 2023:

  • Neither founder could overrule the other

  • Both February 2024 buyout offers were rejected (at the same valuation)

  • There was no neutral arbiter, no independent director, no lead investor to broker peace

  • The only lever left was the Singapore High Court

By the time the court ruled in Hosp's favour in April 2024, AUM had already collapsed from $1B to $80M. The cap table that worked beautifully when both founders agreed had no mechanism for the moment they didn't.

Mistake 2: Anchoring yields to their own house token

 ๐Ÿš€In May 2020, Cake's founders launched DeFiChain (DFI) โ€” a custom Bitcoin-linked layer-1 blockchain. DFI tokens were airdropped to users, and DFI sat at the heart of Cake's yield strategies (Freezer lockups up to 5 years, liquidity mining sometimes >100% APY).

๐Ÿค” At the time, this made sense. Building their own chain felt like vertical integration โ€” full control, transparent on-chain mechanics, a moat against opaque competitors like Celsius and BlockFi.

When crypto winter hit, DFI imploded.

๐Ÿ’ธ From its 2021 peak of ~$4.80, DFI fell to ~$0.045 โ€” a 99% wipeout, against BTC's ~70% drawdown. A different category of pain entirely.

2 things broke at once:

  • Customer portfolios evaporated (much worse than if they'd just held BTC/ETH)

  • Cake's commission revenue cratered, also denominated in DFI

That's why AUM collapsed over 90% from $1B to $80M โ€” far worse than what a generic crypto winter would predict for a custody platform.

๐Ÿง  3 Lessons Learned

Lesson 1: Build governance for the day you disagree

In crypto, founder-controlled cap tables are celebrated โ€” no VC dilution, full upside. But when trust breaks between co-founders, a clean 50/50 split becomes a corporate stalemate. And the only lever left is the courts.

๐ŸŒฎ Key Takeaways:
  • Avoid 50-50 equity splits to prevent decision deadlocks.

  • Independent directors are cheap insurance against future deadlock โ€” one neutral vote breaks a tie.

  • The cheapest moment to install governance is before you ever need it.

๐Ÿ› ๏ธ Operator Playbook:
  • โš–๏ธ Install a deadlock-resolution clause in your shareholder agreement

    • Specify a neutral arbiter (i.e. independent director, mutually agreed mediator) for board-level disputes.

    • Work with a Singapore corporate lawyer (never DIY a deadlock clause).

      • Thatโ€™s why I always recommend my founder friends to work with Sunil (Senior Partner at Dentons Rodyk). He is very well-known in the startup ecosystem and has helped founders navigate costly founder disputes and is an accredited mediator.

      • If youโ€™re starting a company, having founder disputes, fundraising, or doing M&A โ€” you should talk to him.

  • ๐Ÿค Bring in a lead investor or independent director early

    • A small institutional cheque (~10-15%) buys you a board seat that isn't 'you vs your co-founder'.

    • Even one voting director outside the founder block can break a 50/50 stalemate before it reaches the High Court.

Lesson 2: The deeper the vertical integration, the bigger the blast radius

Vertical integration sounds like control.

But the deeper you integrate, the more risks become correlated โ€” token risk, customer risk, revenue risk all collapse into a single line item. When that line item moves, everything moves with it.

๐ŸŒฎ Key Takeaways:
  • In yield platforms, never let stakers, reward payments, and platform fees depend on the same correlated token.

  • Vertical integration concentrates exposure rather than diversifying it.

  • When customer losses and revenue losses share the same trigger, your P&L has no shock absorber.

๐Ÿ› ๏ธ Operator Playbook:
  • ๐Ÿช™ Decouple your platform's economics from any token you issue

    • Take fees in stable assets (USDC, USDT) rather than your house token.

    • Run reward yields through battle-tested external protocols like AaveCompound, or Lido before building proprietary chains.

    • Maintain a treasury split: at least 60% in BTC/ETH/stables, max 40% in any token tied to your own platform.

  • ๐Ÿ“Š Build a "correlated risk" stress test into quarterly board reviews

    • Map every revenue line and customer-facing product to its underlying assets.

    • Ask: "If asset X drops 95%, what happens to AUM, fee revenue, and brand trust simultaneously?"

    • Compare with how Coinbase and Kraken structure custody โ€” no proprietary token in the staking flow, and both survived 2022 with users intact.

Lesson 3: Delayed cost cuts compound into governance crises

After 2021's $631M revenue blowout, Cake committed to costs only sustainable at that level โ€” a $100M venture arm, a SPAC merger, 180 staff.

๐Ÿ“‰ The first operational loss came in Q3 2022, but they waited 14 months to cut headcount. By then, the 30% layoff was so abrupt it broke the co-founder relationship and triggered the winding-up filing.

๐ŸŒฎ Key Takeaways:
  • Plan your cost base on trough-cycle revenue, not peak-cycle revenue.

  • Late cost cuts compound into bigger cuts that fracture internal alignment.

  • The first operational loss is the signal to act. The longer you wait, the deeper the cut.

๐Ÿ› ๏ธ Operator Playbook:
  • ๐Ÿ“‰ Run two financial plans: peak-cycle and trough-cycle

    • Build your operating budget against trough-cycle revenue assumptions

    • Allocate any peak-cycle surplus to the balance sheet (cash, BTC, stables) rather than headcount or office expansion

    • Maintain at least 18 months of runway calculated against trough-cycle revenue

  • ๐Ÿช“ Pre-commit cost actions to operational triggers

    • First operational loss = freeze hiring, freeze new product investment

    • Second consecutive operational loss = layoffs of 10-15% (small enough to preserve team alignment)

    • Coinbase laid off 18% of staff in June 2022 โ€” within weeks of crypto's downturn โ€” and was back to profitability by 2024. Cake waited 14 months and cut 30%.

๐Ÿ”— The Runway Insights

  • The new under-utilised sales trick you can steal (Read)

  • How to use Claude for GTM and pricing analysis (Read)

  • Growing a fully bootstrapped email marketing platform to $37M ARR (Read)

  • I redesigned my landing page so AI agents can read it (Read)

  • The wide path: Why most current valuation dynamics are not here to stay (and what will) (Read)

๐Ÿ’ฐ Southeast Asia Funding Radar

  • Featherless.ai raises $20M (Series A) to expand serverless AI inference platform and build neutral open-source AI infrastructure for developers globally (More)

  • Cata secures $5.3M (Seed) to scale white-label SaaS platform helping F&B and retail brands build owned digital ecosystems and apps (More)

  • FORMAS.AI bags $3.98M (Pre-Seed) to build AI-native design workspace for architecture, engineering, and construction industries across global markets (More)

  • Cube closes $3.7M (Series A) to scale e-commerce data intelligence platform helping leading brands grow faster online across Southeast Asia (More)

๐Ÿ’ƒ Rate Todayโ€™s Edition

What'd you think of today's edition?

Your feedback helps me create better content for you!

Login or Subscribe to participate in polls.

Thatโ€™s all for today

Thanks for reading. I hope you enjoyed today's issue. More than that, I hope youโ€™ve learned some actionable tips to build and grow your business.

You can always write to me by simply replying to this newsletter and we can chat.

See you again next week.

- Admond

Disclaimer: The Runway Ventures content is for informational purposes only. Unless otherwise stated, any opinions expressed above belong solely to the author.

Reply

or to participate.