From Deeptech to Weirdtech
Gil Dibner

Overused, under-specified. Deeptech is a term that gets thrown around constantly but means almost nothing. Investors and founders alike love the term because it sounds complex and defensible — a kind of shorthand for technical moats and hard science — even if it lacks precision. There are vague notions that deeptech means hard tech, but also that "hard tech" refers to hardware (sometimes) but also to difficult engineering or science problems. The term has been so overused it’s become meaningless: a fuzzy, catch-all label — once clever, now just a way of saying “not SaaS” or “there must be a barrier to entry in here somewhere.”

Deeptech, shallow moat. In my view, the term deeptech contains a built-in fallacy: it’s often assumed to imply defensibility, but the opposite is frequently true. When a company’s core concept hinges on a specific technical advance — an algorithm, a model, a material — it can be relatively easy for competitors or incumbents to replicate the innovation and erode margins. The true barriers to entry are rarely in the technical moat alone. They’re often found in product execution, go-to-market strategy, customer relationships, and distribution. Engineering and scientific excellence are important — but they’re only part of the winning formula.

Let's get weird. At Angular Ventures, we’ve come to suspect that “deeptech” just isn’t precise enough. It doesn’t really describe the kind of companies we want to back. We think “weirdtech” might be closer to the mark.

Weird Is a Feature, Not a Bug

Weirdtech companies often look strange at inception. The founders sound a little too intense. The customer doesn’t quite exist yet. The architecture is unconventional. And the pitch doesn’t fit neatly into existing categories. There are no competitors, and it’s unclear that it would ever make sense to start a company in the domain at all. That’s the point. These are the companies that don’t just challenge incumbents — they challenge conventional wisdom and challenge us as investors to throw out our frameworks. The best weirdtech founders aren’t chasing trends. They’re pulling on threads no one else sees — sometimes for years.

The Many Flavors of Weird

Weirdtech comes in many forms:

• Weird markets. These are sectors most investors avoid — either because they seem too hard, too crowded, or too old-school. Think shipping (like Portchain) or custom parts procurement (like Sourcix). These are massive industries with deep inefficiencies, often overlooked because they’re not “cool.”

• Weird technologies. Some companies pursue approaches that go directly against conventional wisdom. These aren’t just novel — they’re heretical. FalkorDB is a good example: it challenges the widely held belief that LLMs alone can underpin knowledge-driven applications, and is delivering a graphRAG solution that is 100x faster than the competition.

• Weird business models. Some of the most exciting bets we’ve made are in categories widely seen as poor fits for venture capital — like nuclear energy, power grid optimization, and space launch systems. We’ve backed companies in all of these areas. They’re weird for good reasons — but if they work, the upside is enormous.

We might be weird enough for you. These companies often get passed over by traditional VCs. They’re too early, too risky, too technical, too hard to explain in a deck. But that’s exactly where we want to be. We fund these bets — not because they’re weird, but because if they’re right, they’re huge. Could it be time to retire the term “deeptech?” It’s lost much of its edge. “Weirdtech” isn’t perfect, but it gets us closer to what matters: the courage to back companies that don’t fit — yet. Companies that seem “bad” in some way or far from obvious. That’s where the biggest outcomes live. If you're building something weird, we want to hear from you.

What if an AI Was Your Best Customer?
Gil explores how AI agents autonomously selecting and consuming tools can reshape the software industry.

Seven Signals
An evolving VC investment decision framework.

Five Emerging VC Swimlanes
The venture market is evolving into five distinct swim lanes, each with its own rules, characteristics, and key players.

Announcing our $125M Fund III 
Announcing Angular Ventures III, a $125M inception-stage fund dedicated to backing technical founders across Europe and Israel solving today’s most complex problems.

ENTERPRISE/TECH NEWS

Meltdown. Sam Altman tweeted “our GPUs are melting” when ChatGPT’s image generation launch was met with user enthusiasm. While X users celebrated the watershed moment in generative AI art by creating their own images in well-known styles (like the Muppets, Renaissance Art and Studio Ghibli), criticism came quickly. Studio Ghibli founder Hayao Miyazaki famously called artificial intelligence "an insult to life itself".

Gemini. Somewhat overshadowed by OpenAI’s image gen and MCP integration news, Google launched Gemini 2.5. “Gemini 2.5 Pro marks a significant leap forward for Google in the foundational model race – not just in benchmarks, but in usability. Based on early experiments, benchmark data, and hands-on developer reactions, it’s a model worth serious attention from enterprise technical decision-makers, particularly those who’ve historically defaulted to OpenAI or Claude for production-grade reasoning.”

HOW TO STARTUP

Rumours. TechCrunch released an expose on 11x last week, claiming that they’d spoken to multiple former employees about issues in the business. The article comes off the back of rumours of a fallout between 11x and their backers, but after Techcrunch published their piece, investors in the start up like Benchmark and A16z came out strongly in support of the company.

Liquidity. Coreweave’s much anticipated IPO ‘missed the mark, raising $1.5B, rather than the targeted $2.7bn. The FT writes that despite the buzz around Coreweave’s positioning as AI infrastructure, there was weak demand from investors and that “another issue involved the Nvidia $250mn anchor order.” “Typically, anchor investors publicly commit early to signal confidence. Nvidia’s last-minute move suggested an emergency patch rather than a strategic endorsement. Given Nvidia’s multiple roles with CoreWeave — shareholder, supplier, customer — its involvement was always going to be viewed as complicated. The belated timing only reinforced concerns.”

HOW TO VENTURE

Network. Rob Hodgkinson writes about Network Centrality in The Signal Rank off the back of the recent Pitchbook report that well-connected investors have superior returns. Taking one facet - ie having a network of other investors - they argue the pros and cons, citing: “The idea that highly networked VCs perform better in boom times (and worse in downturns) is worth considering.” Utilising a network while avoiding group think becomes the challenge.

Firebolt announced its benchmark results on FireScale, a new benchmark simulating real-world workloads. “FireScale shows Firebolt delivering low latency of 120 milliseconds under high concurrency of 2500 QPS and price-performance that is 8x better than Snowflake, 18x better than Redshift and 90x better than BigQuery.”

Budibase was selected as one of the top 10 open source startups in 2025.

Forter made two additions to its executive team, Jim Lejeal as its new CFO and Cyndy Lobb as its CPO.

Fixefy’s CRO Yaniv Butel will be speaking at PSC LIVE Chicago 2025.

Aquant
Senior Enterprise Account Executive (New York)

FalkorDB
Core Developer (Tel Aviv)

Firebolt
Enterprise Account Executive (London)

Sourcix
Head of Marketing (Tel Aviv)