In 2022, VCs poured $1.37 billion on 78 generative AI deals —almost as much as was invested in all of the previous 5 years combined, according to PitchBook data. As I mentioned in my previous posts on creative automation and generative AI, the space is not new, but the pace of innovation is now so fast that it’s hard to keep up with the weekly announcements on new generative AI tools and APIs.
According to a new industry report by Skyquest, AI will contribute $15.7 trillion to the global economy, with $6.6 trillion coming from increased productivity and $9.1 trillion coming from consumer surplus. The global generative AI market was valued at USD 8.12 billion in 2021 and it is expected to reach at USD 63.05 billion by 2028, at a CAGR of 33.7% over the forecast period (2022-2028).
Judging by the current excitement (or hype) in the generative AI space, this is just the beginning. OpenAI recognised this as well and was quick to launch the OpenAI startup fund, a $100M corporate fund (with participation from Microsoft, the main backer of OpenAI and soon to be owner of 49% of it). The fund is expected to make large investments in a small number of companies (not more than 10) where AI can have a profound impact: healthcare, climate tech and education.
The challenges for investors in generative AI
There are a lot of people that don’t get what’s the big deal about Generative AI. It’s great you can create images and text based on a prompt, but isn’t it just a gimmick? Will Fortune 500 companies pay for multiple Generative AI tools to increase productivity? I believe the answer is a resounding yes.
I previously wrote that generative AI will become mainstream when it goes from playful to useful. The launch of ChatGPT (based on GPT-3.5) or Midjourney 4, have taken this technology mainstream and for the first time even risked Google’s dominance in search.
In many ways, we’re just in the beginning. GPT-3 has 175 billion parameters and GPT-4, which is under construction, will have up to 100 trillion parameters. It can pass the bar exam, pass the medical accreditation tests and write a full book on a single prompt.
To get into the investors shoes, it’s important to mention the challenges they face. It’s a bit of a Catch-22. On the one hand, it’s called venture, or risk capital for a reason and venture investors should take risks. On the other hand, mistakes can be costly, especially in a tough market environment like today. So what’s keeping generative AI investors up at night?
Here’s a short list:
- The concentration of power on the API level / Platform dependency – Most Generative AI startups are not developing the core AI tech and are relying on the same APIs for their core technology. Those APIs are controlled by a small number of well funded companies (including OpenAI’s $10 billion in new funding on the way). It will be tough for new startups to compete on AI core technology. They therefore have to focus on the application layer and face serious platform risk. As startups experienced in the past, when the API gets a cold, the startups catch pneumonia.
- Are you backing the right horse? Companies that were doing AI image creation to replace stock photography may have woken up one day to find out that their product was completely commoditised with free products/ APIs. Is there room for a Jasper.ai to be valued at $1 billion when ChatGPT offers something relatively similar for free?
As the generative AI space moves rapidly, it’s tough for investors to know if they’re backing the right horse. For years, investors asked founders “what if Google/Facebook/Amazon copy you?” now that question is trickier, competing not only against the tech giants but also against other startups or established scale ups that are smart enough to add generative AI features to their products. - IP issues, malicious use, future regulatory risk – Midjourney. Stable Diffusion and DevianArt have already been targeted with a class-action lawsuit alleging that allege that these organizations have infringed the rights of “millions of artists” by training their AI tools on five billion images scraped from the web “without the consent of the original artists.”
For generative AI to work, it requires vast amounts of training data, which in most cases is scraped without consent or attribution. I believe we’ll see more regulation in this space.
Like with any new technology, there are risks associated with the use of Generative AI and there will be bad actors who will abuse it. We’ve seen it used for spamming, misinformation, and even for finding vulnerabilities in code.
Regulation (and the platform’s own T&Cs) will have to catch up to the pace of bad actors to quickly address these risks. Demis Hassabis, the founder of Deepmind, is urging we should proceed with caution.
So how can Generative AI startups compete given these platform dependency, competition and IP risks? Is UX enough of a competitive moat? Should they look for riches in niches? There’s no silver bullet, and ultimately startups in this space will have to justify their worth based on unit economics.
The Andreessen Horowitz enterprise team did a great job articulating the fund’s thesis for investing in generative AI:
There don’t appear, today, to be any systemic moats in generative AI. As a first-order approximation, applications lack strong product differentiation because they use similar models; models face unclear long-term differentiation because they are trained on similar datasets with similar architectures; cloud providers lack deep technical differentiation because they run the same GPUs; and even the hardware companies manufacture their chips at the same fabs.
There are, of course, the standard moats: scale moats (“I have or can raise more money than you!”), supply-chain moats (“I have the GPUs, you don’t!”), ecosystem moats (“Everyone uses my software already!”), algorithmic moats (“We’re more clever than you!”), distribution moats (“I already have a sales team and more customers than you!”) and data pipeline moats (“I’ve crawled more of the internet than you!”). But none of these moats tend to be durable over the long term. And it’s too early to tell if strong, direct network effects are taking hold in any layer of the stack.
Based on the available data, it’s just not clear if there will be a long-term, winner-take-all dynamic in generative AI.
Who Owns the Generative AI Platform?by A16Z
A shameless plug for Remagine Ventures
In the generative AI landscape for Israel I published on Jan 13 2023, I listed 50 startups across various categories. I believe there are a lot more coming (some currently choose to be in stealth) and that we will be seeing tools for automation enter every role in the company (coding, sales, marketing, design, etc) across many verticals. We are likely see more specialisation i.e. text for doctors/lawyers, design for commerce/gaming etc.
At Remagine Ventures, we’ve been investing in the media and entertainment tech space since 2019 and have made 3 investments in Israeli startups leveraging generative AI so far: 1) Hour One – text to video using human characters, Piggy – text to mobile presentation, Munch – automated video highlights and overlays. I believe that in a year’s time this landscape is going to be much more crowded and would love to speak with founders building the future automation stack for companies and consumers.
I'm a former general partner at google ventures, head of Google for Entrepreneurs in Europe and founding head of Campus London, Google's first physical hub for startups.
I'm also the founder of Techbikers, a non-profit bringing together the startup ecosystem on cycling challenges in support of Room to Read. Since inception in 2012 we've built 11 schools and 50 libraries in the developing world.
- The potential of Mini Apps to disrupt App Distribution as we know it - December 21, 2024
- Weekly #FIRGUN Newsletter – Dec 20 2024 - December 20, 2024
- 2025 In Media and Entertainment Tech - December 19, 2024
A nice read as always, thanks Eze. I do think UX will play a critical role in this trend, however, I think the real strong moats will come from similar areas as before – network effects, ecosystem (especially with enterprise), etc.
Ultimately, the technology is important, but the success on the business side will depend on fundamentals (CAC< LTV) like every other online business and so I agree with you that the basics, like network effects, will apply here as well.