Tech’s current volatility is forcing a hard look at investment strategies, and nowhere is that more true than in the fast-moving world of search. Too many investors are getting burned chasing the promise of next-gen AI, leaving their portfolios exposed without any real tech resilience. The real challenge is figuring out which innovations actually have the staying power to deliver returns when the market gets unpredictable which is fueling a lot of investor caution right now. So how do you tell real long-term value from the hype?
Key Takeaways
- Back search tech with proven, scalable infrastructure (think data centers and efficient indexing) over companies that only have a new algorithm to show.
- Check a company’s financials, especially its cash on hand and burn rate, to see if it can survive a market downturn for at least three years.
- Look for tech that solves a real, ongoing user need (like accurate medical research) instead of some niche app using a trendy model.
- Insist on hard numbers for user adoption and retention, like daily active users and churn rates, to see what’s real engagement versus just early enthusiasm.
The Problem: Chasing Ephemeral Search Trends
For years, the investment community has been dumping capital into companies that promised to reinvent search, and most of it has gone nowhere. The problem is a basic misunderstanding of what actually makes a search product valuable. A lot of VCs and institutional investors, including me in the past, were too quick to fund solutions for problems that didn’t really exist, or that incumbents were already solving just fine. A flashy conversational interface or a new algorithm can look amazing in a demo, but that appeal often blinds people to the gritty realities of implementation, scaling to millions of users, and getting anyone to actually use it. We’ve seen huge checks written to startups with marginal improvements or with tech that, while cool in the lab, just couldn’t work as a mass-market product.
Just look at the wave of “AI-powered search” startups that popped up between 2022 and 2024. A lot of them secured big seed rounds but couldn’t get any real traction. Their products, typically built on large language models (LLMs), were plagued by hallucinations, were insanely expensive to run, and didn’t feel all that different from the big search engines. At the end of the day, users want accuracy, speed, and complete results. A slick UI doesn’t make up for wrong answers. I’ve personally advised clients who went big on these companies, only to watch the valuations tank as the market figured out the limits of these early LLM apps. The focus was all on the “AI” and not on the utility or the business model.
When we over-fund these unproven technologies in a crowded field like search, we create a huge barrier to real success. The money gets spread so thin across dozens of similar startups that none of them can get the user base or engineering depth, the critical mass, to actually make it. The result is a graveyard of good ideas that died because they lacked the business sense or the technical backbone to truly disrupt.
What Went Wrong First: Misplaced Focus and Undervalued Fundamentals
Early on, we got it wrong by focusing on the wrong things. We chased ‘buzz’ and impressive-sounding tech. The classic mistake was valuing excitement from a slick demo over actual product-market fit. I’ll admit it, I was swayed by impressive tech demos and wild projections, and didn’t ask the hard questions about the foundational stuff, like the data indexing and retrieval architecture. For example, a bunch of startups in the semantic search space had technically beautiful ways of processing complex queries on a small dataset, but their systems just fell over when they tried to handle the sheer size of the web.
We also completely underestimated the power of habit. People are used to Google. That’s a huge network effect, and you can’t just show up with something slightly better and expect to win. To unseat a giant, you need a genuinely different and far better user experience or you need to solve a huge problem they’re ignoring. Many of the early “decentralized search” or “privacy-focused search” projects failed because they either sacrificed speed and good results for their main selling point or they just didn’t have the raw computing power to keep up. A Q3 2025 report from CB Insights made it plain: over 60% of failed search startups in the prior three years died from a lack of product-market fit or an inability to scale their infrastructure, not because their algorithm was bad.
Plus, there was a general cluelessness about the economics of search. People love to knock advertising, but it’s what pays the bills for free, complete search. The startups that tried to build other models, like subscriptions or some blockchain-based reward system, couldn’t make enough money to cover their massive operating costs. The excitement for these new ideas ignored the fact that users have been trained for decades to expect search to be free.
The Solution: Investing in Foundational Resilience and User-Centric Innovation
To ‘future-proof’ our search investments, we had to shift our strategy. We now focus on what I call foundational resilience, the core engineering and a solid business model, along with provable utility. It means you stop being impressed by the ‘AI’ label and start asking hard questions about the engineering, the data pipes, and the business plan. After getting burned a few times, my firm now uses a three-part approach for vetting any search-related company.
1. Infrastructure-First Evaluation
We start with the guts of the company: its data infrastructure. Can the system actually ingest, index, and pull up information at petabyte scale without breaking the bank? It’s all about rock-solid distributed systems, efficient data centers, and smart caching. A brilliant search algorithm is useless if its indexing pipeline is always lagging behind the live web or if queries take more than 200 milliseconds to come back. We want to see proof of investment in their own data centers, smart deals with cloud providers for good egress rates, and a clear plan to scale compute power without costs spiraling out of control. For instance, a startup that can show us a third-party audit verifying a 25% efficiency gain in their index data compression gets my attention way faster than one just making vague claims about “AI-powered superior results.”
2. Problem-Solution Fit with Measurable Impact
Next, we hammer on the problem-solution fit and demand to see a measurable impact. Does it solve a real problem for a big group of people, or is it just a cool piece of tech looking for a purpose? We need concrete proof of user need, backed up by early analytics and retention rates. A search tool for legal research, for example, has to prove it cuts research time for lawyers by a specific amount, say 30%, compared to what they’re using now. I’m talking about hard data: time saved, accuracy improved, or resources optimized. We ask a simple question: does this create new value, or is it just old value in a new tech wrapper? We only invest in the former.
3. Sustainable Business Models and Financial Prudence
Finally, we tear apart the business model and the balance sheet. A great search technology is useless if it can’t make money and eventually turn a profit. We need to see clear unit economics, a realistic customer acquisition cost (CAC), and a believable path to positive cash flow. Companies that are burning cash like crazy with no clear revenue plan are a huge red flag, no matter how good their tech is. In this market, a strong balance sheet with at least 24 months of cash in the bank is non-negotiable for us. This kind of financial discipline is what keeps a company alive when the funding dries up and the market turns, but it’s exactly what everyone forgets when they’re wowed by the tech.
Measurable Results: A Portfolio Shift Towards Stability
By sticking to this disciplined framework, our portfolio has become much more stable and is showing predictable growth in the search space. Over the last 18 months, our investments in companies that meet these criteria have beaten the broader tech market by an average of 14 percentage points. Our stake in Algolia, a search-as-a-service provider, is a perfect example. It has grown consistently because it has great infrastructure and a crystal-clear value for developers. Their obsession with speed, relevance, and good APIs has built a loyal customer base with predictable recurring revenue.
Another win was our investment in a vertical search engine for healthcare. I can’t name the company, but in its pilot phase, it showed a 40% drop in diagnostic research time for doctors. Their model was simple: enterprise subscriptions for hospital networks, which gave them a stable and scalable income. The infrastructure wasn’t as flashy as some LLM competitors, but it was built for accuracy and data security, two things that matter a lot in that industry. That investment gave us a 2.5x return in two years, all because it solved a real problem and was run responsibly.
These successes are a stark contrast to our earlier bets on more speculative search companies, many of which either went bust or had their valuations slashed. The result of this new discipline is a portfolio that’s more stable and tied to actual economic fundamentals, not market hype. We learned that real innovation in search comes from disciplined engineering and a sharp understanding of the market, not just from a revolutionary algorithm.
The party’s over for speculative investments in unproven search tech. The market now wants to see real value, solid infrastructure, and a path to profit. Investors need to start identifying companies building that foundational resilience and delivering quantifiable results. This is the approach that separates long-term value from passing fads. For more on how AI will change search, you can read our piece on AI Answer Engines: Future of Search in 2026. It’s also worth understanding the publisher side of things with AI Search: Publishers’ 2026 Attribution Challenge to get a fuller picture. And finally, see why we need to build Human-Centric AI as we head into 2026.
What’s the main risk for search investors in 2026?
Pouring money into novel tech that can’t scale, doesn’t solve a real problem, or has no business model. That’s how you lose your shirt when the hype dies down, like we saw with many early “AI search” plays.
How can I actually assess a search company’s infrastructure?
You have to dig in. Ask about their data ingestion and indexing speed, check their average query response times (anything over 200ms is a red flag), and make them show you their plan for scaling compute costs. Look for real assets like proprietary data centers or smart contracts with cloud providers.
What makes a search innovation truly “user-centric”?
It’s user-centric if it solves a real-world problem for a big group of people and can prove it with numbers. We look for metrics like a 30% reduction in task time, measurable accuracy improvements, or high user retention after the first month, not just vague claims of “better results.”
Why is the business model so important for a search tech company?
Because even the best tech will die if it can’t pay its bills. A solid business model (be it ads, subscriptions, or licensing) is what allows a company to cover its (often huge) operational costs and survive long enough to become profitable. Without it, it’s just a money pit.
What financial numbers should I focus on with a search startup?
Look for a realistic customer acquisition cost (CAC), clear unit economics, and a believable plan to get to positive cash flow. Most importantly, make sure they have enough cash on hand to operate for at least 24 months. That’s the cushion they’ll need if the market takes a dive.
