I was interviewed by TheMarker (Hebrew only) on why Israeli startups haven’t been part of the mega funding rounds in AI. I thought it’s a good timing to elaborate on where I see opportunities for Israeli startups in this rapidly growing space.
To put things in context, global venture funding in Q2 2023 reached $65 billion, an 18% decline quarter over quarter, and a 49% drop compared to the second quarter of 2022, when startup investors spent $127 billion, according to Crunchbase.
In comparison, Israeli venture capital investments declined even further in the first half of 2023, to $3.2 billion, a 73% YoY drop, according to a new report by Viola.
Amidst this bleak funding environment, AI has been a saviour, reports the WSJ. Companies categorized as AI in Crunchbase raised $25 billion in the first half of 2023, representing 18% of global funding. That includes the $10 billion round into OpenAI, the $1.3 billion round into inflection as well as smaller rounds, like the $105M seed round into 4-week old Mistral, a French startup developing LLMs to rival OpenAI.
Most of the companies that raised these mega rounds are relatively early stage in terms of product and monetisation (if at all) and about half of the companies that raised large sums, are developing LLMs, which generally requires deep pockets to cover 1) cloud costs, 2) expensive deep learning engineers 3) get the data to train the models. That’s not to say that application layer startups didn’t raise large funding rounds: Jasper ($125M), RunwayML ($141M) Tome ($43M), ElevenLabs, But Israeli startups, to a large extent, have been excluded from this trend, only securing smaller round.
The state of Generative AI startups in Israel
According to The Global AI Index by Tortoise Media, the first index to benchmark nations on their level of investment, innovation and implementation of artificial intelligence, Israel is the 7th most attractive AI hub globally.
To date, AI21 Labs is the only Israeli startup developing LLMs (it raised an impressive $64M in July 2022), and most of the companies operating in this space are in the application layer or the tooling/infrastructure layer. Another Israeli founded startup that raised significant amounts is Pinecone ($100M series B in April), a vector database for data scientists.
In the last version of the Israeli generative AI landscape that we, Remagine Ventures, published in April 2023, we listed about 71 startups. We’re in the process of updating the landscape (you can add a company via this form or by scanning the QR code) and have identified at least 40 new startups in this space where generative AI is the company’s main value proposition/ tech stack, not including all the incumbents like Wix or Lemonade, who have started incorporating generative AI features into their products.
But these companies are entering an increasingly competitive market, where fundraising has become tough. Are they too late?
It’s not easy for investors to allocate in this space
Back in January, I published “Investing in Generative AI startups” in an effort to share my considerations allocating capital in this space as a venture investor. In a nutshell, there are a lot of challenges in the space and investors are being picky. Six months later, the speed of change has nothing but accelerated and the complexity increased.
However, as Dawn Capital pointed out in their recent thought piece on the rise of generative AI in Europe, there are two important mitigating factors for investors:
- The speed at which businesses and users can create AI applications has been turned on its head
- The cost to quality ratio of outcomes has reduced dramatically
Investors who dismiss generative AI companies and jump to the conclusion that there are ‘no investable opportunities for startups in generative AI’ are taking the lazy way out which supports their decision to stay on the sidelines. As John Luttig writes in his excellent post ‘Hallucinations in AI‘, while it’s unreasonable to expect that generative AI will change everything overnight, there are many opportunities for startups. I tried to consolidate my thoughts as well as other perspectives from leading venture capital funds.
Opportunities for startups in generative AI
For many in the tech industry, the current wave of VC investments in AI companies represents a new bubble. Others talk about the scale of the AI revolution as a massive platform shift as big as the introduction of the iPhone. Regardless of where you stand on this debate, there’s no denying that generative AI presents a huge opportunity in terms of economic value. $4.4 trillion dollars a year. That’s the economic impact of generative AI per year, according to new research by McKinsey. Will startups reap the fruits of this new trend, or is it the battle of giants?
AI Vertical SaaS – As mentioned by Index Ventures in a recent post, there’s a rise in AI native SaaS companies, building automations for specific industries. I’m already seeing interesting teams tackling the adtech/marketing tech stack, and could see this expanding to many other verticals where Israel has strong talent including cybersecurity, developer tools, fintech, retail tech and commerce, etc.
Workflow automation – the reason Jasper.ai is still in business despite ChatGPT’s popularity is the fact they embedded their product (content copyright for marketing) into their clients workflow. Marketers who are creating copy for campaigns appreciate having everything plugged in so they can take the AI generated copy and easily put into action with the tools they’re used to working with. I can imagine there are opportunities here across industries/verticals.
Neural search and vector databases – as mentioned in my previous post on VC Cafe, search is being transformed by AI. Vector search (like Pinecone) will become the preferred way to store and manage unstructured data. While the consumer space tech is dominated by Google, the enterprise vector search, utilising the organisations own data and applying the power of AI, remains an attractive opportunity for startups.
Autonomous agents – while technologies like AutoGPT or BabyGPT are still not mature (as I wrote on my post on AgentGPT), but their goal-seeking attributes make them an attractive interest area for technical founders. Imagine autonomous agents that help a brand reduce churn, increase basket size, post social content for maximum impact, etc.
Building on top of the Open Source generative AI stack – Open source LLMs are constantly increasing in quality and are likely to continue get better. For example, MosaicML released MPT-308, an open-source model that is free for commercial use and outperforms GPT-3. This trend can dramatically reduce the costs for startups, reduce the reliance on big companies and create a market for tooling and infrastructure companies to build on top of the multi-model stack. While closed LLMs like OpenAI, Anthropic and Cohere will continue to have an edge in terms of quality, open source LLMs will continue to grow in popularity.
Multi modal LLMs: today, the solutions in the market are primarily text, image, video or voice based. We’re starting to see the first multi modal AI agents, where users are able to make an input in one modal, say voice, and get the output in another. One example of this is Imagebind by Meta. Early adopters in this space can come up with interesting use case that aren’t currently being addressed by ChatGPT or Bard.
Generative AI for media and entertainment. Generative AI has the potential to revolutionise media and entertainment, by creating new forms of content, personalising the user experience, and improving the efficiency of production. I covered this in my post on generative AI in gaming, and my partner Kevin Baxpehler wrote about the impact of generative AI in media and entertainment in this VC Cafe post.
If you’re an Israeli founder building in this space at the pre-seed level, please don’t hesitate to get in touch.