Hey — It’s Nico.

Welcome to another Failory edition. This issue takes 5 minutes to read.

These are the 5 most important things:

• Zing, an HSBC backed fintech, shut down — learn why below.

• Data suggests 2025 will be full of failed startups.

• OpenAI launched Operator, an independent AI agent.

• Retro Biosciences is raising $1 billion to extend human life.

• DeepSeek is everywhere — get the full breakdown below.

Let’s get into it.

Save 10 hours a week by hiring an Executive Assistant.

MEET DEVNI: The Right Hand to the CEO of Morning Brew

Austin’s Executive Assistant saves him 10+ hours a week by handling his emails, calendar, travel, meeting prep, and more. This gives him the time to focus on what matters most—growing his business.

Most outsourced firms can’t source the type of EA who can work with a $100M executive, but Oceans can.

Work with the firm that sources the right-hand to Morning Brew, Beehiiv, and Magic Spoon.

Get full-time, highly experienced talent for only $3,000/month.

Learn more about Oceans

This Week In Startups

🔗 Resources

Data suggests 2025 will be full of failed startups.

Making Markets in Time

The Post DeepSeek R1 World

How to find your Activation Metric

📰 News

Perplexity makes a bid to merge with TikTok.

OpenAI launched Operator, an independent AI agent.

Anthropic is raising $1B from Google.

SoftBank might invest 25 billion on OpenAI.

💸 Fundraising

Retro Biosciences is raising $1 billion to extend human life.

UVeye raised $191 million to develop an “MRI for Cars”.

Heliom, a nuclear fusion startup, raised $425 million.

Castelion raises $100 million for Hypersonic Strike Weapons

Fail(St)ory

HSBC’s Fintech Flop

In January 2024, HSBC launched Zing, an international money app designed to make cross-border transactions seamless. Just a year later, it’s gone.

Despite a $150 million investment and the backing of one of the world’s largest banks, Zing couldn’t carve out a place in the fintech landscape. It wasn’t just about funding or infrastructure—corporate-backed startups often struggle to compete with more agile, independent players. Zing is yet another example of how even deep pockets can’t guarantee success in a fast-moving market.

What Was Zing: Zing was HSBC’s attempt to enter the fintech market, aiming to compete with established platforms like Revolut and Wise.

Launched in January 2024, the app featured a user-friendly interface, enabling customers to manage multiple currency wallets, perform quick bank transfers, and make international payments with transparent fees and competitive exchange rates.

The app Zing also offered a Visa debit card, which could be added to digital wallets like Apple Pay and Google Pay, allowing for seamless spending and ATM withdrawals globally.  

Despite these features, Zing struggled to differentiate itself from competitors. While it provided services similar to those of Revolut and Wise, it didn’t offer significant innovations or unique value propositions to attract users away from these established platforms. Additionally, being part of HSBC, Zing faced challenges in agility and adaptability, which are crucial in the rapidly evolving fintech landscape.  

In theory, Zing had the resources and infrastructure to challenge its competitors. However, in practice, it failed to deliver anything substantially different from what was already available in the market.

The Numbers:

• 📅 Launched: January 2024.

• 💰 Investment: $150M from HSBC.

• 🧑‍💼 Team Size: 400 employees affected by the closure.

• 📉 Lifespan: Operated for just 12 months before shutting down.

Reasons for Failure: 

• Lack of Differentiation: Zing entered a crowded market dominated by Wise and Revolut, both of which had already established themselves as leaders in international money transfers. Unlike its competitors, Zing didn’t offer a compelling value proposition to stand out. The app’s features were all things users could already do elsewhere, often more easily or at lower costs.

• Corporate Prioritization: In August 2024, HSBC underwent a significant leadership change. The new CEO led a shift towards cost efficiency and operational improvements. This change in leadership led to a decreased focus on Zing, ultimately resulting in the decision to shut down the app. 

• Slow Response to Market Demands: Startups like Wise and Revolut thrive on their ability to iterate quickly, learn from user feedback, and adapt their products. Zing, tied to the bureaucracy of a global bank, didn’t have that flexibility. By the time it hit the market, it was already behind.

• Burning Cash Without Traction: While $150 million might seem substantial, it pales in comparison to the funding secured by competitors. Revolut has raised approximately $1.99 billion over 13 funding rounds, with a valuation reaching $45 billion. Similarly, Wise has garnered around $1.03 billion across 15 funding rounds. Zing had the resources to launch but didn’t find product-market fit quickly enough to justify further spending.

Why It Matters: 

• Corporate priorities shift fast. A change in HSBC’s leadership turned Zing from a priority into a liability overnight. If your startup depends on a corporate parent, you’re always one meeting away from shutdown.

• Product-market fit is everything. Without it, even $150M isn’t enough. Zing had infrastructure, a solid team, and deep pockets—but none of that mattered because users didn’t need it.

• Customers don’t care who backs you. HSBC’s reputation didn’t make Zing a success. Users pick products based on value, not the size of the company behind them.

Trend

DeepSeek

This week has been all about DeepSeek—everyone seems to be talking about it. So today, let’s break down why this matters, what the consequences might be, and how this could actually be good news for AI and AI-related startups.

What Happened?

On January 10, DeepSeek released DeepSeek-V3, a general-purpose model similar to OpenAI’s GPT-4. It’s designed for standard AI tasks like text generation, summarization, and general language understanding.

Then, this week, everything exploded with the release of DeepSeek R1, an open-source chain-of-thought model.

What’s special about R1? It performs similarly to OpenAI-o1, but at a fraction of the cost. Unlike traditional AI models that require massive computational power, R1 delivers strong results using significantly fewer resources.

And here’s the kicker: It’s open-source.

Why It Matters

• It’s surprisingly efficient. DeepSeek built a high-performing model without the kind of insane compute power that OpenAI, Google, and Anthropic rely on. That alone makes it notable.

• It’s open-source. Instead of locking it behind a paywall, DeepSeek made R1 (and their other models) public. That means anyone can use, tweak, and build on it—something the big players are hesitant to do.

• It shook the market. NVIDIA’s stock took a hit because this model’s efficiency suggests companies might not need to buy quite as many expensive GPUs to train strong AI models. That got investors nervous.

• It proves China is a real AI contender. The US has been restricting AI chip exports to slow China’s AI progress. If models like R1 can thrive without high-end chips, those restrictions might not be as effective as hoped.

The NVIDIA Stock Collapse

One of the most significant side effects of DeepSeek’s launch? NVIDIA’s stock took a big hit.

In fact, NVIDIA experienced the largest single-day loss in market value by any public company in history, shedding $593 billion as its stock plummeted 17% to $118.42.

Why? Because if AI models can be built more efficiently, that could mean less demand for NVIDIA’s ultra-high-end GPUs. Investors didn’t love that idea.

Does this mean NVIDIA is in trouble? Not really. The biggest AI companies are still going to buy absurd amounts of GPUs. But if efficient AI models like R1 become a trend, the rate of NVIDIA’s growth could slow down.

Is This Good News?

The past few days have been full of strong opinions on DeepSeek.

Some see this as a major threat to American dominance in AI. Others raise concerns about censorship, given that DeepSeek operates under Chinese regulations.

But on the other hand:

• DeepSeek is open-source. Anyone can run it, including researchers and startups outside of China. In fact, Peplexity has already integrated the model to their platform and removed the censorship.

• It doesn’t require supercomputers. Unlike other top AI models, R1 can run on just a couple of GPUs.

• It democratizes AI development. This lowers the barrier for researchers and smaller companies to build and experiment with AI models.

• It boosts competition in AI. A more open, global AI space could lead to faster innovation and more diverse perspectives in AI research.

• Startups could benefit. Companies that currently rely on GPT models may cut costs by experimenting with DeepSeek instead.

Help Me Improve Failory

That's all of this edition.

Cheers,

Nico