What if Sequoia, Goldman and Barclays are all wrong about the AI bubble?

Making the case for AI: The First Riskless Tech Revolution

We are living through one of the most interesting technological revolutions in last 50 years. AI is going to change how we work, how we communicate, and even how we socialize.

I am not much for estimating these kind of revolutions but if you need a number — both McKinsey & Accenture just put a $10 trillion tag on the impact of AI on global GDP by 2028.

Recently, Sequoia Capital wrote a memo though talking about a potential bubble in AI and the “missing $600B” in revenue. And then, Goldman Sachs wrote a report GenAI: Too much spend, too little benefit?

But I am here to tell you how AI is a very different kind of bubble when you look at these revolutions and bubbles through the lens of risk and returns.

All three prior revolutions — internet, cloud, and mobile — are very similar to this AI transformation where the economic impact is in trillions of dollars.

But where they differ is in how much risk early players took to get us there.

Two kinds of risk: tech risk & market risk

In very simple terms, here is how define the two risks:

Tech risk refers to the risk of whether we know how to build something.

For example, when you build the very first electric car or a giant space shuttle, there’s a lot of things we simply don’t know and therefore there’s a ton of tech risk. The very first transistor, the very first mainframe, the very first iPhone model all had ton of technology risk — we didn’t know whether we could build those things till we built them. Lots of unknowns.

Market risk refers to the risk of whether we know how to make money from something.

For example, when you build the very first electric car — you don’t know if anyone will buy it? if so how many people? at what price? what will it cost to build one? to ship a million?

This time it’s different. Here’s why:

Now, let’s look at internet, mobile and cloud — and then AI.

Internet had tech risk and market risk

While it looks so obvious today, it took years — nearly a decade plus for us to figure out how and what the internet was — whether it would be built on phone lines or cable, what kind of spectrum, what kind of modems, what kind of networking protocols, how much bandwidth, what speed — all of it was up in the air till we grappled with it and the internet started looking mostly like what it looks like today.

This means internet had high tech risk.

We also didn’t know how, or even if, we could make any money with or on the internet. People who are generally very smart about tech and business — like Bill Gates — wrote books on the internet and got almost all of it wrong. (See “The Information Superhighway”). Very core concepts that generate most money on the internet today like display and link advertising were invented after years and years, and even after they became the norm almost no one could have predicted the multi hundred billion dollar revenues for companies like Facebook and Google simply from running ads.

Cloud had minimal tech risk but high market risk

Companies like NetSuite and Salesforce emerged by 1999, and most people could see how to take on-premise apps and rebuild them for the internet. While people may have disagreed on exact shape of this move, no one thought it was impossible. In short — minimal to no tech risk.

However, even the smartest business people — who made 10s of billions of dollars in the prior era in tech — could not figure out if and how profits would be generated in the cloud.

This is why Larry Ellison and Bill Gates & Steve Ballmer led companies almost completely ignored the cloud. They were convinced there was little to no profits to be made — as proven by lack of profits at Salesforce & NetSuite even a decade into their existence.

In short, cloud had high market risk.

Mobile apps had minimal tech risk but high market risk

Just like the cloud, once you see an iPhone with a few apps, you can somewhat easily see how apps could be written for the phone even if you disagree on things like programming languages and such.

Low to no tech risk.

But almost no one knew how to make money off of it — till likes of Uber, Spotify & Instagram — showed us the path with transactions, subscriptions, and ads.

In short, there was market risk.

AI is fundamentally different

Today, we know two things: AI can do amazing things that most never imagined even just a few years ago, and the newer models from multiple players keep showing us more and more potential.

In short, we know how to build models — and not just one company, we have now several companies building better and better models almost every month.

We went from high tech risk just 5 years ago — will AI ever really be able to do certain things — to GPT2, GPT3, GPT4 — and Mistral, and Llama — the pace of releases of models of better and better quality at lower and lower costs is collapsing the tech risk down to extremely low.

On the business potential and business model — how much money can we make and how to charge for it — we know that the path to monetizing AI is pretty much same as our business models for internet, cloud and mobile:

• Ad revenue: just like search companies, charge by views or clicks.

• Subscription revenue: just like Salesforce and Spotify, charge per user per month.

• Infrastructure revenue: just like AWS and Azure, charge for usage.

The AI revolution is the first revolution where we both know how to build it, and how to monetize it.

You can argue whether the winners will be companies like AWS, ServiceNow and Microsoft that will sell AI to existing customers, or OpenAI, Perplexity and Anthropic that will sell AI to new customers — but the business model is clear.

What does lack of business and tech risk mean for AI?

This framework leads to a few predictions. Let’s start with the impact on venture capital.

VC as an asset class makes money because they are willing to take a risk when others won’t and get paid for it. When VCs invested in likes of Yahoo! and Amazon, they were taking on a lot of risk. Similarly, angels and VCs that invested in Salesforce — which found it very hard to raise capital — were also taking on a lot of risk, as were investors in Uber.

But, today, companies like OpenAI and Anthropic find it easy to raise capital — because even the CIOs of largest banks and insurance companies already believe AI is going to be hugely impactful to how they run their business.

This is very different from the last 3 revolutions — the smartest business leaders who ran the best run companies of that time mostly ridiculed or ignored them because it was not clear if they would work and if they worked it was not clear if they would make any money.

Today, people like Jamie Dimon want to be leaders in AI — these companies were all laggards when it came to the internet, mobile, and cloud.

The valuations of Anthropic and OpenAI reflect the obviousness of the bet.

The founders and these VCs have to now go build things assuming AI is the “obvious bet”, and this is why you see VCs like Sequoia writing about the $600B gap, and VCs like Foundation Capital writing about “After LLMs”.

As a founder, I love this — we are living in a unique time in history where one of the largest tech revolutions is “obvious”, and we get to all assume its going to happen — and go build tools and apps on this new platform.

I am building the data security & privacy layer for AI, and we are betting our company on it.

What are you building or investing in assuming AI revolution is low risk and inevitable?

________

Sources:

McKinsey report (https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/beyond-the-hype-capturing-the-potential-of-ai-and-gen-ai-in-tmt#)

Accenture report (https://www.accenture.com/us-en/insights/consulting/gen-ai-talent)