Amber CEO Michael Wu: After the FTX Collapse, I Went Through "the Biggest Setback of My Life"

In this episode of the “Friends of OKX” interview series, we sit down with Amber Group CEO Michael Wu.

Michael looks back at Amber Group’s journey — from its rapid expansion in 2021 to the existential threats and hard lessons learned during the FTX and Three Arrows Capital (3AC) fallout in 2022. He shares what drove the team’s early pivot from TradFi quant trading to going all-in on crypto, breaks down the core mechanics of market making, and explains why strict compliance is essential for surviving tail risks.Looking ahead, he explores how blockchain’s decentralized and permissionless nature makes it the ideal underlying payment and settlement network for a future AI agent-driven economy. Finally, Michael shares his personal advice on managing founder stress, along with his core criteria for investing in and

The audio transcription is done by AI and may contain errors.

From Morgan Stanley to Crypto

Mercy: Michael, thanks so much for joining me today. Do you usually speak more Chinese or English day-to-day?

Michael Wu: A bit of both, actually. I live in Singapore now, but I went abroad when I was 15 and did my undergrad in the US.

Mercy: So you went straight into TradFi right after graduation?

Michael Wu: Yeah, I joined Morgan Stanley. They were actually on the 37th floor of this very building. Our current office is just above them.

Mercy: Did you intentionally set up shop right above your old employer?

Michael Wu: No, no, it wasn’t that deliberate! We were just looking to move offices and found a good space here. It is a funny coincidence, though, because my very first job right out of college was here in the ICC (International Commerce Centre).

Amber’s Turning Point

Mercy: We were just talking about how you’ve been a founder for nine years. When the 3AC collapse happened in 2022, what year were you into your startup journey?

Michael Wu: About our fifth year.

Mercy: Facing something like that in your fifth year must have been a massive hit.

Michael Wu: 3AC affected us indirectly, but the real blow was FTX at the end of the year. Amber was my first startup. We started prepping in late 2017, officially launched in Q1 2018, and were lucky enough to hit profitability by Q2.

2021 was our hyper-growth phase. We rode the wave of crypto’s first real institutionalization, where clients and investors started treating Bitcoin and crypto as long-term strategic assets. We experienced explosive growth, peaking at around 1,300 employees globally. We closed major funding rounds — bringing in Temasek and Brunei’s sovereign wealth fund in late 2021 — and hit a $3 billion valuation.

Then 2022 arrived, and the bull market turned bearish. The 3AC and FTX blowups were massive, historic shocks to both our company and the entire industry. I remember in H1, we were preparing for our Pre-IPO with two major investment banks. We got our first term sheet at an $8 billion valuation, but I pushed back, saying we were worth at least $10 billion. Barely a week later, the cascading liquidity crises and blowups began.

Overnight, we went from a soon-to-be IPO company with a potential $10 billion valuation to a completely dried-up funding market. Q1 2022 marked our five-year anniversary, and by Q2, we recorded our first-ever quarterly loss.

Things worsened in H2. The FTX collapse caused us massive direct financial losses. We were forced to aggressively downsize and cut business lines. We grinded through the entire 2023 bear market and have slowly gotten back on track. But looking back, I believe these hardships are an invaluable asset for both me as a founder and Amber as a company.

The Dual Tests of Life and Entrepreneurship

Mercy: What was your immediate reaction when the crisis hit? It seems like your life had been pretty smooth sailing up until then. Was this your first major setback?

Michael Wu: You could definitely say that. My co-founders and I had similar backgrounds — family, school, early careers — everything had gone relatively smoothly without any major hurdles. So the FTX collapse in late 2022 was probably the biggest setback of my life, not just my career. Looking back, a lot of it still feels surreal.

On top of the FTX blowout, my co-founder unexpectedly resigned at the end of that year, which was a massive blow to me. Throughout late 2022 and early 2023, we were entirely focused on firefighting and getting the company back on track, which actually kept me moving. But I remember by late 2023, once the dust had mostly settled, I sank into a pretty deep slump.

It’s human nature to think linearly — we expect life to just move in a straight line, or even accelerate upwards. Taking a massive hit for the first time and having to pick ourselves back up was a completely new experience. It was a profound lesson for me personally, and for our young company as well.

Company Survival First

Mercy: Looking back, what was the toughest — yet ultimately the best — decision you made during that crisis?

Michael Wu: It’s hard to call any of them the “best,” because so many were just incredibly painful. For example, downsizing and being forced to let go of so many teammates. My absolute top priority at the time was simply making sure the company survived. We had to stay alive; we had to keep pushing forward.

If there was one “correct” thing we did, it was holding fast to our core identity. We constantly reminded ourselves of who we are: an ambitious, tech-driven firm that always strives to stay at the cutting edge of innovation.

Decisiveness Through Bull and Bear Markets

Mercy: You must have been under unimaginable pressure at the time. How did you manage to stay calm and make those tough calls?

Michael Wu: Honestly, as a founder, you actually tend to be calmer and find decision-making easier in those moments. During extreme hardships, your options are drastically reduced. Once you’ve thought things through and weighed the pros and cons, the choices you have to make — and how you execute them — become incredibly urgent and clear-cut.

It’s actually much harder during a bull market or when the industry is expanding rapidly because there are simply too many choices. That’s when you really need to slow down and think things through. It might look like you have a ton of great opportunities, but the truly right choices are always limited.

Back to your question: while late 2022 was incredibly challenging, there wasn’t a whole lot we could do. We just had to pick the best path out of a few limited options and execute it to the best of our ability. The formula was simple: we needed capital, we needed our core team, and we had to protect our core business and clients.

Mercy: Your mindset actually reminds me a lot of our boss [Star Xu]. What you just mentioned about navigating through bull and bear markets while staying product- and tech-driven — he thinks the exact same way.

Michael Wu: Yeah, I have a ton of respect for Star Xu. He’s not just an OG in our industry; he is a highly ambitious entrepreneur with a genuine vision for building great products.

From Quant Trader to Crypto Builder

Mercy: What exactly drew you to crypto? What made you decide to quit such a glamorous job and go all-in on Web3?

Michael Wu: I have to admit, my initial leap into entrepreneurship and leaving Morgan Stanley wasn’t actually because of crypto. When we first started, we weren’t in this fancy ICC building. We were in an old, rundown building, in a two-person office right above a seafood stall. Back then, we were actually called Amber AI.

The spark for starting the company actually came from my time as an FX and Rates trader at Morgan Stanley. It was my first job out of college. In your first couple of years at an investment bank, your main job is literally buying breakfast and coffee for everyone. You weren’t allowed to take notes; you had to memorize the orders for a desk of a dozen people. Then, you’d pull up a stool next to the senior guys and just watch how they traded.

But around my second year, I was given my first real assignment: managing Morgan Stanley’s Asia EFX (Electronic Foreign Exchange) Book, overseeing its algorithmic pricing. I had a background in programming — I started coding when I was 10 — so my boss handed this to me. It was actually a crucial role because, back then, most pricing was still done via “voice trading.” It was very old-school; traders would quote prices manually or over the phone.

They gave the EFX book to me because Asian currencies weren’t mainstream yet, and the volume was small. But from 2014 until I left in 2017, in just four short years, electronic pricing volume completely surpassed traditional quoting.

Those were the years when the internet was booming, and digital payments and neo-banks were emerging. Yet, the traditional finance sector was still moving at a very slow, archaic pace. In 2017, seeing Google DeepMind’s AlphaGo play Go was a massive shock to me. I realized that technologies like machine learning and AI were inevitably going to completely disrupt and revolutionize the entire financial industry.

So that’s what we set out to do when we founded the company in 2017. We were probably one of the first startups in Asia to use machine learning and neural networks for pricing, liquidity management, risk control, and trading. A lot of the big firms didn’t dare touch it because these algorithmic models were considered “black boxes.” You couldn’t easily explain the process, so they didn’t want the liability. As a startup, we didn’t have that baggage; we were eager to innovate.

Mercy: So how did you transition into crypto?

Michael Wu: Also in 2017. When we started, we were trading multiple asset classes. But within two or three months, we realized that 95% of our revenue and profit was coming from just one asset class: Crypto.

I remember around the Chinese New Year in 2018, I spent a few days discussing it with my co-founders. We were a team of about ten people at the time, and we came to the conclusion: we go all-in on crypto.

Coincidentally, by Q1 2018, the market had already transitioned from bull to bear. Bitcoin had dropped maybe 50%, and altcoins like Ethereum had plummeted by 80–90%. But our flagship offering at the time was a market-neutral asset management product. The market was still highly inefficient back then, so there were massive arbitrage opportunities.

I remember on my birthday in 2018, Bloomberg Tech ran a headline along the lines of, “While the market crashes, some institutions are still highly profitable” — and they were talking about us.

That instantly put us on the map. Given our backgrounds, we already knew a lot of family offices and former TradFi bosses. They approached us wanting to invest in Bitcoin and crypto, so we rode that wave and quickly built out a business catering to institutions, HNWIs, and family offices — similar to private banking in traditional finance.

This remains our core business today. Even our current Nasdaq-listed entity (ticker: AMBR) and platforms like WhaleFin grew from this exact foundation.

We were definitely lucky in the early days; it was often about being in the right place at the right time. Interestingly enough, I first came across Bitcoin back in 2015 through my job. When I was a trader at Morgan Stanley, my boss didn’t understand it at all and asked me to look into it. I realized the concept was actually viable. At the time, Bitcoin was only hovering around $1,000.

Rethinking Blitzscaling and True Long-Termism

Mercy: What was your biggest takeaway from that experience? Are there decisions you make today specifically because of what you went through back then?

Michael Wu: Absolutely. The most obvious one is embracing true long-termism. Before 2022, my so-called “long-termism” was a bit hypocritical. What do I mean by that? I always believed we were building something monumental, riding generational tech waves like crypto and AI that would leave a mark on history. But because our early years were so smooth, even though our vision was long-term, our execution was incredibly rushed.

Back in 2021, I was obsessed with a book called Blitzscaling. It uses this metaphor that a startup is like throwing yourself off a cliff and assembling an airplane on the way down. The premise is explosive growth — expanding at all costs to achieve network effects and dominate the market. I bought both English and Chinese copies and handed them out to all our executives.

It wasn’t until the 2022 crash that I realized, while there’s truth in that playbook, a lot of it just doesn’t apply to us. Blitzscaling works for the Web2 era, where scaling eventually drives marginal costs down to zero and maximizes network effects. Crypto, however, is inherently financial. Many opportunities come from migrating heavily regulated financial models onto this new technology stack. That reality completely clashes with the “blitzscaling” mentality. I’ve had to really reflect on that logic and realize it doesn’t fit our journey.

Coming back to long-termism: I still believe our mission is massive, something I might spend my entire life building. And because of that, I no longer care if we achieve it tomorrow, next year, or in ten years.

Jensen Huang recently said something in an interview that deeply resonated with me: “I have 50 years to change the world; I don’t need to do it tomorrow.”

That really hits home for me now — maybe I’m just getting a bit older. Looking back, one of my biggest strategic mistakes in 2022 was benchmarking against the wrong targets. I rarely compare myself to others, but that year, my mindset shifted. Fueled by the Blitzscaling hype, I abandoned true long-term thinking.

We expanded rapidly and blindly. My mandate at the time was a new product feature every two weeks and a major iteration every month. Metaverse, NFTs, crypto payment cards — if I thought of it, we had to build it or buy it. I was so impatient. Of course, the whole industry was overly euphoric and rushed at the time, but looking back, we definitely lost our way. Surviving that phase taught me that as long as we eventually realize our vision, there is no need to rush. That is my biggest area of personal growth.

I’m a true believer in crypto, but my convictions differ from many of my peers. Right now, people looking at crypto applications generally fall into three camps:

Extreme Financial Nihilists: They see this as a giant casino, purely for making money.

The Backend Believers: They think crypto will fade into the background and become the underlying settlement layer for TradFi. (I actually agree with this, but I think crypto is much more than that.)

The Tech Utopians: Usually newcomers or pure idealists who believe in absolute decentralization — not just DeFi, but decentralized social media, governance, everything. I think that’s a bit too unrealistic.

My perspective sits somewhere else. Regarding the financial aspect of crypto, I believe it pushes permissionless access to the absolute limit. TradFi will eventually plug into this because of those permissionless benefits, and this massive regulatory migration will create tremendous commercial opportunities.

As for non-financial crypto applications, I believe they hold immense value — not necessarily for humans, but for the future Agent Economy. Once AI agents, rather than human individuals, become the primary economic participants, crypto’s core traits will be vital.

Until that happens, the trade-offs just aren’t worth it for the average human user. Think about it: most people won’t sacrifice convenience, user experience, and cost just to use a clunky, expensive service simply because it’s “decentralized and trustless.” Humans want fast, cheap, and ideally free.

But if you are an AI agent, your calculus is completely different. Cost matters, but user experience and “convenience” are irrelevant to code. What does matter is:

Permissionless Access: An AI agent can’t afford to wait for human approval to execute a task.

Trustlessness: Relying on a third party introduces uncontrollable variables that AI cannot predict.

Decentralization: An AI’s processes won’t be arbitrarily interrupted by human intervention.

Suddenly, the three biggest friction points of blockchain applications become massive advantages for AI agents, and the drawbacks become completely acceptable. Once the economy is driven by AI agents — which I believe is only 3 to 5 years away — non-financial blockchain applications will unlock explosive, unprecedented value.

Envisioning the AI Agent Economy

Mercy: What would an AI agent-driven economy actually look like in practice?

Michael Wu: We’re actually starting to see glimpses of it already. When we hail a ride or watch short videos, an algorithm is doing the matching and recommending in the background. The difference is that, right now, these programs don’t make their own commercial decisions. They just execute predetermined business goals set by the corporate entities behind them.

But imagine giving an AI agent its own wallet. Let it price its own services and store its value using tokens, and let it handle payments directly with humans or other agents. At that point, the agent becomes an autonomous decision-maker.

What does this future economy look like? A company essentially becomes a network of agents, using crypto-priced economic resources to determine the most optimal decisions.

These agents will then dispatch services. Some will be online, like the apps we use today. Some will be offline and physical, executed by robots. And don’t just limit your thinking to humanoid robots — a robot could be an autonomous cart clearing tables in a restaurant, or even a door handle with a small chip running a localized, on-device model. That is an AI agent. For every service and utility provided, the agent will collect data, execute the task, and process the payment.

Eventually, I believe all applications will be agent-driven at the top layer, and powered by crypto at the base layer.

This shift is already underway. We have drone deliveries, and Tesla’s FSD (Full Self-Driving) is getting incredibly good. If a large portion of cars become fully autonomous, car owners won’t just own a vehicle for personal transport; they will own a yield-generating asset. You aren’t just employing a digital driver; you own an asset that creates economic value, whether you are using it yourself or renting its utility to others.

In this kind of economy, payments will absolutely be programmatic, and we will see a massive volume of high-frequency micro-transactions or “streaming payments.”

Humans actually hate the psychology of streaming payments. Take streaming media today — you often don’t finish a show, you just watch a few minutes and switch. A decade ago, people asked, “Why not just charge by the minute?” The answer is that humans hate the mental pressure of thinking, “Every minute I watch costs me money.” We’d much rather pay a flat monthly subscription and watch whatever we want.

But when you own an autonomous asset, your personal experience is completely detached from its usage. The only goal is maximizing utility and yield. Traditional payment networks simply cannot handle this kind of fragmented, continuous micro-transaction volume. When that time comes, crypto will be the only viable solution.

How Market Makers Provide Liquidity

Mercy: Since you mentioned your background as an FX trader earlier — which obviously ties into what Amber does today — could you break down exactly what a market maker is and what they actually do?

Michael Wu: Amber is widely known as one of the largest and most successful market makers in crypto, though that’s just one of our business lines. At its core, market making is simply liquidity provision.

What does that mean? When someone wants to buy an asset, there needs to be an ask price; when they want to sell, there needs to be a bid price. Ideally, there is always a price available so you can execute a trade the exact moment you want to. The entity providing those bid and ask prices is the market maker.

Anyone can be a market maker — even a retail user placing a limit order on an exchange. But institutional market makers like us use technology to do this at a very high frequency and provide massive liquidity depth. We ensure that when someone wants to trade an asset, they can do so cheaply and efficiently.

The execution is straightforward: we use proprietary algorithms and AI to populate the order book, keeping the bid-ask spread as tight as possible to minimize trading costs for everyone, while keeping the order book as thick as possible.

There are two golden rules in our line of work:

Liquidity begets liquidity. Liquidity is most expensive when you need it the most. During extreme volatility, true price discovery is incredibly difficult. As a market maker, if someone sells an asset to you, you might not be able to offload it; if someone buys from you, you might not be able to buy it back. That’s why liquidity often dries up completely during extreme market events.

I can proudly say that surviving so many extreme market cycles has only strengthened our reputation. People realize that, compared to others, we consistently step up to provide liquidity and do everything in our power to honor our commitments to the market.

Why Compliance is Crucial

Mercy: I know Amber places a heavy emphasis on compliance. Beyond just checking the boxes for regulators, why do you think compliance is so crucial?

Michael Wu: That’s a great question. I’m not a compliance expert myself, so I can only share my personal perspective — this doesn’t necessarily represent Amber as a whole.

Why is it important? First off, it’s for your own survival. It’s about mitigating compliance risks, which we treat as “tail risks.” You can choose to ignore them because, most of the time, they don’t blow up. But the moment they do, they can completely kill a company or cause catastrophic damage.

In our industry, compliance is often incredibly difficult and full of gray areas. People either think the worst won’t happen, or they’re willing to gamble that it won’t. But when it does, the consequences — whether it’s a massive fine or a total wipeout — are fatal. We’ve always been a heavily risk-focused company, and this is exactly like cybersecurity: if you don’t invest in security, one hack can wipe you out. Compliance works the exact same way.

Because we genuinely believe in true long-termism and have a massive vision for what we’re building, we have to invest heavily in compliance. Zooming out a bit: as economic activity and the AI agent economy migrate on-chain, the sheer scale of the financial interests and societal impact will be staggering. Governments and regulators will inevitably step in. The earlier we lay the groundwork for this, the stronger our moat becomes. When that paradigm shift finally hits, our compliance infrastructure will be the exact advantage that allows us to operate while others cannot.

Second, compliance protects the entire industry. If one or two companies skirt the rules, they might just hurt themselves. But if it happens on a massive scale, it shatters regulatory confidence and inflicts severe negative shocks on the whole space.

As a leading player, we share a consensus with our peers: we can’t let the industry get ruined. We at least need to do our part and do it right. Plus, we lived through 2023, and that period was brutally painful.

I remember a time when no one wanted to touch crypto. It became a complete taboo. Our clients are family offices and institutions; we have great relationships, and they kept their accounts open with us. But back in 2023, if we brought up new crypto opportunities, they simply refused to look. It wasn’t until bullish catalysts like the spot ETF approvals that confidence gradually returned. The regulatory cycle shifted from a headwind to a tailwind, finally giving the industry a chance to move forward again.

Navigating Compliance: Client Insights and the Future

Mercy: I’m noticing a lot of shared values here — like committing to true long-termism and fully embracing compliance. OKX is also heavily focused on the compliant route. Have you noticed any shifts in your client demographics or their primary concerns since you decided to go all-in on compliance?

Michael Wu: In the early days, “compliance” didn’t really exist in crypto. It was almost impossible because the industry was just too new. But our fundamental approach hasn’t changed much, simply because our core clients have always cared about risk. Since we mostly deal with institutions and family offices, the people managing that money aren’t going to risk their careers dealing with an unregulated entity. If something blows up, they’re the ones held responsible.

Of course, this path comes with its own headaches. For instance, we’re licensed in Singapore and will likely secure licenses in other jurisdictions soon. Sometimes, strict regulatory frameworks force us to migrate clients between regional entities. Naturally, clients get frustrated. They’ll say, “I’ve already onboarded with you once! Why do I have to submit all my paperwork again just to migrate to a new region?”

We actually tried launching a retail business back in 2022. It didn’t pan out for various market reasons, so we wound it down later that year. But looking to the future, I believe AI agents will completely revolutionize this space by offering highly tailored services to every single client, big or small — especially for on-chain users.

This will be a game-changer for user onboarding. Imagine an AI agent instantly analyzing your source of funds, capital size, intended use cases, and interaction patterns. Based on that data, it can dynamically determine the precise level of KYC (Know Your Customer) or KYB (Know Your Business) you actually need. Our incubator actually backed a project tackling this exact problem earlier this year, and I think it’s a brilliant direction for the industry.

Managing Anxiety and Information Overload

Mercy: First off, as a CEO, you’re processing a massive volume of information every single day. When you’re dealing with high uncertainty or intense pressure, how do you filter the noise and manage anxiety?

Michael Wu: I’m actually someone who replies to messages very quickly. For the vast majority of it, I try to process and respond within 5 to 10 seconds. But if it’s something crucial, I’ll add it to my to-do list and let it marinate. It’s a habit left over from my days as a trader. I often do these solo “workshops” — basically blocking out dedicated time just for myself to deep-dive into specific topics. As for the anxiety? It’s honestly not a big issue; I generally don’t get that anxious.

Mercy: So you’re just a total optimist, then?

Michael Wu: Yeah, I’d say I’m pretty optimistic. There are definitely times when the pressure gets intense, but as long as I can still get some sleep, I’m good.

Investment Logic and Looking Ahead

Mercy: I know Amber is actively investing and running an incubator right now. When you’re evaluating a project, what is the absolute most important thing you look for?

Michael Wu: The team is definitely crucial; I always want to sit down and actually talk with them. Personally, my interest always leans toward the intersection of Crypto and AI. Even if a project isn’t building directly in both, if I can see a clear, well-defined role for it within my broader vision of an AI-driven, crypto-native future, I’m highly attracted to it.

Mercy: On the flip side, is there an absolute dealbreaker? A red flag that guarantees you won’t invest?

Michael Wu: The founder’s character. In our industry, a “rug pull” can sometimes be hard to strictly define. But if I interact with someone and get the sense that they are fundamentally a bad actor, or if their core values seem completely skewed, I generally stay far away from them.

Mercy: Just a quick, fun one to wrap up: What is one item that best represents you, or something you just really like?

Michael Wu: Honestly, probably a black T-shirt.

Mercy: And what is your personal resolution for the New Year?

Michael Wu: Definitely to stay true to our original mission and keep leading the company to achieve our ultimate vision.

Follow us
Twitter: https://twitter.com/WuBlockchain
Telegram: https://t.me/wublockchainenglish