We're halfway through 2024 — how should we invest in AI applications?

We're now halfway through 2024, and there's been no shortage of chatter about AI application investing. The Linear Capital investment team gets asked about its views on AI application investing all the time, as well as about progress on Linear Bolt, the dedicated AI application investment program it launched — Linear Bolt is Linear Capital's vehicle for early-stage, globally oriented AI applications. It carries forward Linear's investment philosophy, focusing on projects where technology drives transformative change, and aims to help founders find the shortest path to their goals. Whether in speed of action or investment approach, Bolt's promise is lighter, faster, and more flexible. For a collection of past Bolt articles, click here.

Recently, Can Zheng, Managing Director at Linear Capital and head of Linear Bolt, sat down with GeekPark's "Founder Park" to answer these questions. Below is the full interview:

Key takeaways:

1. What you spend on R&D for tomorrow has nothing to do with today's revenue. Infrastructure depreciation should be viewed over a longer time horizon.

2. Whether it's a large enterprise or a startup, if you can make money after adopting AI, that signals a real possibility of maturation.

3. Even veteran revolutionaries face new problems. Experienced founders need to stay vigilant against letting experience become baggage.

1. AI Project Maturity Jumps Significantly Every Six Months

Q1: What's Linear Capital's investment strategy for H2 2024? Any changes?

Can Zheng: Linear Capital is actively embracing the AI track overall, so there won't be major changes to our second-half strategy. As the AI industry matures, we may get to see some large-scale applications that have already validated PMF. From our investment perspective, we still need to adjust our strategy in response to industry developments, moderately increase check sizes, and potentially invest at slightly later stages.

We actually started observing AI applications in the first half of 2023, but the market was still in its infancy then. Most founders were still in exploration mode and hadn't fully figured things out, so we didn't invest aggressively at that stage. By late 2023 and early 2024, we found some more mature AI application projects. At the same time, more and more founders with deep domain expertise began thinking seriously about AI's possibilities, and their ideas became more mature. We believe change in AI happens extremely fast — project maturity jumps significantly every six months, and investor strategies shift considerably as well.

Q2: Looking back at this wave of AI investing, what has Linear Capital gotten right? What hasn't quite panned out?

Can Zheng: This AI wave is still in very early days. If I had to point to something we got wrong, I'd say that in 2022 and the first half of 2023, we were too eager to see mature founders and mature products. We invested a lot of energy in that expectation, but in retrospect, founders also need time to let their thinking settle and their ideas mature.

Q3: How do you view the AI M&A wave overseas in the first half of this year?

Can Zheng: For early-stage investing, it's definitely a positive signal. Acquisitions are a very common market behavior in the US, and in a sense, this creates a relatively healthy growth environment for both founders and investors.

It also reminds us that we still need to figure out PMF. Even after finding a company's PMF, you might still get acquired — but that kind of acquisition carries far more value.

Q4: What impact does the tightened IPO environment have on investors?

Can Zheng: In the US, the average cycle from founding to IPO is about 11 years. In China, that cycle may be even longer. So for funds focused on early-stage investing today, talking about IPOs right now is a bit like marking the boat to find the sword.

The current IPO environment will have to change. If it doesn't, what happens to all the tech companies built over the past seven or eight years? There are quite a few good ones in there. If they can't go public, these tech companies could face a grim situation. I hope we'll see some changes in the next two to three years.

2. In a Sense, the Inflection Point Has Already Happened

Q5: Many funds talk about investing in AI applications, but in practice they're still quite cautious. What might be behind this?

Can Zheng: Investment institutions are being cautious for a few reasons. On one hand, the industry's maturity level still isn't there — the sector is in very early stages. And after large models emerged, founders and developers still need considerable time to learn AI, understand AI, and integrate new technology with existing industries. That takes a long time. On the other hand, AI itself is still changing rapidly, and this rapid technological change imposes new learning costs on developers.

From model release to AI actually being usable in industry is a very long process. The entire ecosystem — including middleware, infrastructure, and other critical links — is gradually maturing and co-evolving with applications. Without support from these critical components, application development becomes harder and slower. Everyone needs to understand one thing: what I spend on R&D for tomorrow has nothing to do with today's revenue. Infrastructure depreciation should be viewed over a longer time horizon.

Whether it's a large enterprise or a startup, if you can make money after adopting AI, that signals a real possibility of maturation. In a sense, the inflection point has already happened. Take a product like ChatGPT — it's already generating revenue. ChatGPT launched in 2022, did $1.6 billion in revenue in 2023, and is projected at $3 billion for 2024. Looking at this product alone, it's likely profitable. That's clearly a validated PMF. Then you have mature SaaS companies like ServiceNow seeing increased revenue and profit because of AI. The massive capital flowing into AI today is being stockpiled for next-generation technology and products. Today's high spending is meant to bring about the next inflection point.

Q6: How do you view the different types of AI entrepreneurs in the market?

Can Zheng: AI, as an emerging field, has interaction patterns fundamentally different from traditional ones. Experienced founders have many natural advantages — greater resilience, richer cognitive experience, stronger team management skills. But even veteran revolutionaries face new problems. AI carries extremely high uncertainty, and experienced founders need to stay vigilant against letting experience become baggage. There are also many young founders today. They may lack experience in team management and go-to-market, but they have sharper instincts for new technology.

Today's AI still lacks a lot of scaffolding and middleware, so its benefits for large teams aren't that obvious yet. But for small teams, the benefits are very clear. In a way, this actually narrows the gap between young founders and experienced ones — which is a good thing for young founders.

When screening Bolt projects, we place more emphasis on founders' deep market understanding, rigorous experimental validation, and the project's value potential. What we value most are founders' learning ability and execution ability. Execution ability has two main components: one, can you iterate on product quickly; two, can you execute according to your own plans, making rapid trials and iterations between product and market.

3. Launching Bolt to Specifically Target Early-Stage, Global-Market AI Applications

Q7: Why did Linear Capital launch Bolt in H1 2024?

Can Zheng: First, given that large model-driven AI projects are still in their infancy, the startups in this space are similarly early-stage. AI itself is still early, so Linear Capital established the Bolt investment program specifically for early-stage, globally oriented AI applications.

Second, AI has also brought major changes to the startup ecosystem. AI-era founding teams don't need to be large — they can get products out very quickly and find their PMF through rapid iteration. These projects may not need massive capital early on; what they need more is fast investment decision-making. Bolt itself is designed to connect with early, fast-moving projects and provide early-stage support to founders.

Third, today's AI founders have a global outlook. Many projects are going global from day one. Bolt was created to invest faster, lighter, and more flexibly in these early-stage projects with global potential.

Finally, AI-era founders come with global thinking and are good at using AI to become so-called super-individuals. But they may lack experience in fundraising, market expansion, and internationalization. So they may need help in these areas. Bolt doesn't just provide capital — it also produces content and organizes events to comprehensively support young founders, giving them a platform to launch quickly and accelerate growth.

Q8: Are Bolt's investments riskier?

Can Zheng: We spent a long time thinking and discussing this, but ultimately reached a clear and firm consensus: AI is a domain we cannot avoid and should actively embrace. As firsthand witnesses to the AI technology transformation and early adopters of AI products, we can deeply feel how AI boosts productivity. We can see the value AI creates today, and we can fully imagine the changes AI will bring across different industries and domains in the future. So we must embrace AI and find the best way to invest in it.

If we're talking about any single early-stage project in isolation, yes — the earlier the stage, the higher the risk. But this high-risk characteristic is typically already reflected appropriately in valuation.

Q9: What share of Linear's total capital goes to AI projects?

Can Zheng: Internally we maintain flexible allocation across investment domains. Beyond Bolt, Linear's main fund also invests in AI. Bolt focuses mainly on early-stage AI applications. Precisely because it's early-stage, it represents a relatively small share of total capital. If projects funded by Bolt show strong momentum later on, we can follow on at more reasonable and advantageous prices.

Q10: What market feedback and results has Bolt received since launch?

Can Zheng: In the first half of the year, we screened and evaluated numerous projects, ultimately investing in roughly 7-8 AI applications spanning education, job search, companionship, content, consumer, and smart hardware. Bolt is aimed at global-market AI applications — there are no restrictions on scope.