The AI bubble is real

Every day I get a newsletter from Sifted (you should sign up, if you haven’t). And it’s so hard to ignore the buzz around artificial intelligence (AI).

From the rapid adoption of ChatGPT (I literally use it for everything now!) to so many AI startups raising millions, sometimes it feels like I’m living in a golden age of innovation!

But with every gold rush I always ask: Is this sustainable growth or just another bubble waiting to burst?

If I take a step back, I think there are a lot of parallels to past tech bubbles, like the dot-com crash and the crypto craze. And trust me, I’ve I lost thousands 😢 investing in crypto.

Right now, I believe valuations on AI startups are ridiculous and detached from company fundamentals. 

My definition for a "Tech Bubble"

History shows us that "bubbles" form when speculation overtakes rational decision making in the investment world.

Usually I look out for three things:

1. Unsustainable growth: Companies that grow too quickly without a clear path to profitability.

2. Overvaluations: Valuations far exceed intrinsic value or earnings potential.

3. Herd mentality: Investors flock to the trend, fearing they’ll miss out (FOMO).

AI today checks all these boxes as startups. Here’s an example:

Inflection AI raised $1.5 billion in June 2023, giving it a valuation of $4 billion.

Yes, this was founded by Reid Hoffman, so most would argue it has some clout. But a $4 billion start-up, only 1 year since founding, sounds extreme. And according to one investor, Inflection AI has virtually no revenue!

If you look at the public markets, which usually lag the private markets, Big Tech companies like Nvidia have seen their stock prices skyrocket. Nvidia’s market cap went past $1 trillion in May 2023, driven largely by hypothetical demand for AI chips. Nvidia's recent PE ratio has also been 50+, suggesting significant over valuation.

So immediately I have my guard up.

Hype vs. Reality

The Hype

The excitement around AI often feels overblown.

Without naming names, every week I see companies adding “AI” to their branding to attract investment, much like “dot com” was tacked onto company names in the late 1990s.

In fact, a study by PitchBook in 2023 showed that the number of startups incorporating AI in their pitch decks grew by 78% year-over-year, yet many have no clear monetization strategy.

One example of hype is the wave of generative AI tools, with countless startups launching ChatGPT alternatives.

While some bring innovation, many are indistinguishable from existing solutions. I believe the sheer volume of undifferentiated offerings is a warning sign of froth.

The Reality

Don’t get me wrong, I am a sceptic, but through the noise, there are always startups that shine through. For example:

• In Healthcare: AI tools like PathAI are improving diagnostics, while DeepMind’s AlphaFold solved a 50-year-old problem in protein folding, revolutionizing drug discovery. Viz.ai, are also accelerating stroke detection and treatment, saving lives and reducing healthcare costs.

• In Finance: Companies like Stripe are using AI to enhance fraud detection, reducing chargebacks and saving billions for businesses.

• In Supply Chain: ClearMetal leverages AI to optimize inventory management, cutting costs and improving delivery times.

• In Enterprise Software: Tools like Gong.io use AI to analyze sales calls, providing actionable insights that improve close rates by over 30%.

These examples do tell me that AI is (and can) have real-world impact.

What does this mean for investors? You need to spot the right opportunities in AI.

Overall I believe the investment community is placing too much emphasis on the "proprietary" technology AI companies claim to build, with little emphasis on what makes a start up a real business.

My approach to picking winners will be to start with company fundamentals:

1. Revenue Growth: Looking for startups with consistent, high margin revenue streams. Or start-ups with immediate plans to generate revenue within the next quarter. False promises of eventual revenue are risky. In fact if the founders say they will be able to generate revenue in the quarter, I'll be tempted to wait, see if it's been executed and then invest, even if it means at a slight premium.

2. R&D Spending vs. Results: High R&D is good, but it must translate into tangible advancements or defensible IP. Start-ups with masses of funding without any results years down the line is a serious red flag.

3. Customer Stickiness: If revenue isn't the focus right now, someone needs to be using their product. In which case, strong retention rates indicate a product’s value and scalability. But that isn’t enough, if someone loves a product, they would be immediately telling others too. So, what would their new customer referral rate?

I believe focusing on these areas will help weed out the hype from the reality.

And then ONLY AFTER THAT, would I get into the technology itself.

Specifically, great AI companies should have:

1. Proprietary Data: Access to unique, expansive, and domain-relevant datasets is a cornerstone of competitive advantage. Tesla, for example, collects vast amounts of real-world driving data from millions of vehicles, enabling it to train models for autonomous driving. This data not only improves performance of the model but also creates significant barriers for competitors. My question would be: Does the company possess exclusive data that is both large-scale and directly relevant to its applications?

2. Model Innovation (if applicable): Companies often chase breakthroughs in model architecture, training methods, and deployment strategies, but true innovation isn’t just about creating something new, it’s about doing it smarter. I think this means improving efficiency, whether it’s through sparse models or better computational resources. So, the real question is: Is the company pushing the envelope not only in building advanced models but also in maximising training efficiency and reducing resource waste? That’s where the game changing potential lies.

3. Real World Application: A strong foundation in cutting edge techniques and their practical application is essential DeepMind, for instance, integrates reinforcement learning and neural networks, but has a domain specific approach. This means it can address problems like protein folding, once considered unsolvable. My question would be does the company have the interdisciplinary expertise to translate AI breakthroughs into solving a real world problem? If not, then this company is going no where. The world doesn't need any more "AI marketing copywriting tools"

What does this mean for health tech founders? Keep building

Yes, we’re in a bubble.

AI is everywhere, and valuations are soaring. But as a founder, your job is to cut through the noise and keep building.

A few things to keep in mind:

• Solve a real problem: AI won’t save a weak business. If you’re solving a genuine customer pain point, the tech you use is secondary. Customers don’t care about AI, they care about outcomes and solutions. If they’re willing to pay, that’s the only validation you need.

• Stay disciplined on valuation: When VCs are throwing money at AI startups, it’s easy to get caught up in the hype. But be smart. Does your valuation make sense relative to your actual revenue? Overpricing yourself now could make your next round, or survival, much harder.

AI is a tool, not a business model.

Focus on real impact, real customers, and real traction.

The rest is just noise. 🚀