There is a strange mood in venture right now. I was talking to the managing partner of a billion dollar fund, who told me he has yet to make a deal in 2026. He’s not alone.
On one hand, AI is producing some of the fastest-growing companies in history. On the other, many software founders are hearing a quieter version of the same question in every investor meeting:
What stops OpenAI, Anthropic, Google, Microsoft, or the next open-source model from doing this?
That question has become the new cold shower. It sits behind the polite feedback. It explains why perfectly good SaaS companies can suddenly feel unfundable. It is also why deeptech, infrastructure, robotics, defence, semiconductors, energy, and anything with a physical or scientific barrier suddenly feels more attractive. Deeptech has become venture’s emotional safe space. I wrote about this in
The logic is understandable. If AI makes code dramatically cheaper to write, then code itself is no longer a moat. If a small team can “vibe code” a credible product in a weekend, then the old SaaS playbook of shipping faster, designing better workflows, and selling seats into a known category is under pressure. Baillie Gifford put it well:
“AI is collapsing the cost and time required to build software,” but the key point is that this is “not destroying software, but redistributing where value sits.”
That redistribution is what investors are trying to price.
The public markets have already reacted. In April, Bloomberg reported that the iShares Expanded Tech-Software Sector ETF had dropped more than 27% in 2026, while a SaaS index was down almost 40%, driven by fears that AI would permanently weaken demand for software. Venture has followed the sentiment. Crunchbase data shows AI captured close to 50% of global venture funding in 2025, and in Q1 2026 AI companies attracted $242 billion, roughly 80% of all global venture funding that quarter.
No wonder founders feel the bar has moved.
But the conclusion that “only deeptech is fundable” is too simplistic.
What is becoming unfundable is not software. It is thin software.
A thin workflow layer. A dashboard over someone else’s data. A prettier interface for a problem that can be solved by a model. A generic “AI copilot for X” with no proprietary wedge, no distribution advantage, no workflow ownership, and no reason to exist once the underlying model improves.
That is a very different statement from saying software is dead.
In its 2026 Software x AI report Sapphire Ventures correctly pointed out:
“AI is now both the primary bull and primary bear case for software.” The same force compressing valuations for exposed incumbents is also expanding the opportunity for companies that rebuild products, pricing, and business models around AI”.
The monday.com example is a useful counterpoint.
monday.com was supposed to be one of the poster children for software commoditisation. Work management, project tracking, internal workflows, dashboards, automations. These are exactly the kinds of categories people assume AI agents will replicate or bypass.
And yet, after reporting Q1 2026 results, monday.com shares reportedly jumped as much as 26% in pre-market trading. The company delivered $351.3 million in revenue, up 24% year over year, raised guidance, reported record operating income, and launched what it calls an AI Work Platform with native agents.
The lesson is not that every SaaS company can rebrand as AI and be fine. The lesson is that investors are still willing to reward software when AI makes the business more strategic rather than less defensible.
monday.com is not just bolting on a chatbot. It is trying to become the orchestration layer where humans, workflows, data, and AI agents meet. It is also moving toward a “seats plus credits” model, which matters because per-seat pricing becomes fragile when one employee can operate multiple agents. The better question becomes: how much work is the platform actually doing?
That is the shift Foundation Capital predicted when it wrote that AI companies would “sell actual work completion rather than just workflow enablement.”
This is where the software moat debate gets more interesting.
In the pre-AI SaaS era, a moat could be product velocity, UX, workflow depth, integrations, sales execution, brand, switching costs, or a growing dataset. In the AI era, those things still matter, but only if they compound into something the model layer cannot easily absorb.
The strongest software companies will have at least one of five advantages.
They own a system of record. They are embedded in a mission-critical workflow. They have proprietary data or context. They control distribution. Or they can price against outcomes, transactions, consumption, or work completed rather than seats alone.
That is why some software companies may become more valuable in the AI era, not less. AI agents need trusted data. They need permissions. They need audit trails. They need integrations. They need a place where work is assigned, executed, checked, and recorded. In other words, they still need software.
Andreessen Horowitz made the optimistic version of this argument recently: “There will still be moats, and as long as there are moats there’s plenty of reason to expect hugely successful and highly durable businesses to survive and thrive.”
That sentence is important because it pushes back on the lazy version of the AI narrative.
AI is not making all software worthless. It is making weak software obvious.
This is especially relevant for early-stage founders. In 2021, a good-looking SaaS product in a growing category could raise because the market believed in cloud adoption, seat expansion, and the repeatability of the SaaS model. In 2026, that is not enough. Investors want to understand why the company becomes stronger as AI improves.
That means founders need sharper answers to these questions:
- Why does your product become more useful when models get better?
- What data or workflow do you accumulate that others cannot access?
- What action do you take on behalf of the customer?
- What outcome are you willing to own?
- What happens if the UI disappears and the user interacts through an agent?
- Do you become the agent, the tool the agent uses, or the system that verifies what the agent did?
These are not cosmetic questions. They are fundability questions.
For Israeli founders, there is an obvious temptation to lean into what Israel already does well: cyber, defence, semiconductors, infrastructure, robotics, industrial AI, and applied deeptech. That is rational. The world is rewarding hard things. Physical constraints, security requirements, regulatory complexity, and technical depth can all create real defensibility.
But it would be a mistake to conclude that application software is over.
Some of the most interesting companies in this cycle will be software companies that look different from SaaS. They may sell outcomes instead of seats. They may combine software and services. They may use agents to collapse the labour cost of a whole business process. They may start narrow, own a workflow deeply, and expand into a system of action.
Bessemer has described AI as “the biggest wave of technology change we’ve ever seen.” The implication is not that founders should stop building software. It is that founders should stop building software as if the SaaS rules of the last decade still apply unchanged.
The bar has moved.
The old question was: can you build it?
The new question is: why will it matter once everyone else can build it too?
Deeptech answers that question with physics, science, capital intensity, regulation, or proprietary IP. Great software companies will answer it with workflow ownership, data, distribution, trust, and AI-native business models.
So no, it is not true that nothing is fundable apart from deeptech.
But it is true that “just software” is no longer enough.
The next great software companies will not be protected by code. They will be protected by context, execution, customer love, data gravity, and the ability to turn AI from a feature into a business model.
That is a much higher bar.
And probably a healthier one.
He is a former General Partner at Google Ventures (GV) in Europe, former head of Google for Entrepreneurs in Europe, and founding head of Campus London, Google's first startup hub. Eze writes on Israeli tech, venture capital, artificial intelligence, and founder strategy.
He is also the founder of Techbikers, a nonprofit that brings together the startup ecosystem on cycling challenges in support of Room to Read.
- Is Anything Fundable Anymore, Apart From Deeptech? - May 14, 2026
- Tiny Episodes, Big Business: The Israeli Startups Betting on Micro-Dramas - May 12, 2026
- Weekly Firgun Newsletter – May 8 2026 - May 8, 2026
