Welcome back to Innovation Download, our monthly look at the news impacting Australia's startup scene.

This month we’re talking about the SaaSpocalypse. It's dramatic, but has become the go-to phrase to describe the sudden turn against traditional SaaS companies in the public markets, and it had a clear trigger. Anthropic dropped a set of Claude features that just do the work for you, data analysis, expense tracking, the stuff that used to need a paid seat, all in plain language. Investors responded by going into full “voting machine” mode, as the Warren Buffett quote goes (ps. here's a great read on Mr Market). And the selling has come in waves, one Claude release at a time. Claude Cowork, launched in January, and on its own took double digits off Adobe and Salesforce. Then the security tools landed: Claude Code Security pulled around 5% off the cybersecurity sector in February, and in March, word of a model that finds software vulnerabilities on its own, Claude Mythos, sent CrowdStrike and Palo Alto Networks down roughly 6% in a single day.

So in this issue we dig into what’s causing the voting machine to go into overdrive and why it may head back to being a “weighing machine”. Hint: it’s unlikely SaaS is dead.

Let’s kick things off with a video, and who better than Nvidia's Jensen Huang, the man arming the AI companies with their weapons of choice, GPUs. He has a view which you might not expect.

▶️ Nvidia CEO Jensen Huang on AI's pressure on software stocks (CNBC)

In the clip, Jensen Huang argues the market has overreacted. As you might expect, he thinks the disruption is real, just nowhere near as deep as the selloff might suggest, and points to the companies using the tools that are apparently going to destroy them. He goes on to say that what is more likely than collapse is that the incumbents adjust, and his example is ServiceNow (the most prehistoric of all enterprise companies): his view is that data is the game, and the players that have it are in the driver’s seat.

Obviously, Jensen is selling the GPU firearms that the AI upstarts and the incumbents will be using in their battle for customer supremacy, so he wins either way. However, the fact that he thinks the incumbents aren’t dead in the water is likely reassuring to many in their C-suites – or at the very least, they can forward the video to their boards.

Hit play, then read on.

This Month's Reads

Is the share market headed toward a ‘SaaS-pocalypse’ – and what would that mean? (The Guardian)

Rundown: This is the piece that put a name to the panic. The Guardian lays out how a US software selloff became the “SaaS-pocalypse,” and why it travelled so fast: if an LLM can do the job, why keep paying for bespoke software seat by seat. The same local names from up top all got caught in it, though they’ve clawed back a fair bit in the last few weeks. The line that sticks is Morningstar’s, in a somewhat reductionist tone: if one person can now do the work of two, seat counts fall.

The Takeaway: "AI makes software redundant" is what many will take from this, and you can see why. The FUD writes itself: digital photography killed Kodak and Netflix marked the end of Blockbuster. But a selloff, as I noted above, is the voting machine doing what it does.

We saw it with the GameStop short squeeze in 2021, when the meme stock rocketed on nothing but a Reddit narrative. The “wisdom of the crowds” isn’t always wisdom, sometimes it’s just sheep following each other sheep.

As the build part gets easier, the other parts remain hard. Maintaining and improving software is still difficult. Granted it’s easier with agents running around making PRs, but we’re still not at the point where the whole stack is self-driving. That day may come, but do you really want to build, maintain and improve your own HR or payroll software? Maybe just leave that to ServiceNow. Which is what I think the vast majority of companies will do.

Rethinking the SaaS Apocalypse (Forbes)

Rundown: Steve Banker, who covers supply chain and enterprise software for Forbes, reckons the selloff that wiped hundreds of billions in software value is a massive overreaction. His own first take was that pure workflow apps were exposed and optimisation tools safer, until a LinkedIn piece by ReverseLogix CMO Chuck Fuerst changed his mind. Fuerst's argument: AI made the easy part fast, writing code, which creates a dangerous illusion that building software was ever the hard part. It isn't. AI handles the visible layer, the interface and the first version. It doesn't touch architecture, reliability, integration, compliance or maintenance, and most internal builds don't fail at version one. They fail at everything after, as APIs change and what worked six months ago quietly breaks. That's why companies aren't ripping out core systems for AI-built replacements. They're extending the platforms they already trust, not rebuilding them.

The Takeaway: AI did collapse the cost of building, and the weekend-build stories are real. But building was never the business. As Fuerst puts it, the maintenance, the compliance, the integrations that break at 2am aren't edge cases, they are the operating environment. Version one is the easy part. Version two is where companies quietly go to die, and AI only made the first one free. The value moved to the part AI can't carry. An agent is brilliant at processing information but still needs a filing cabinet, and the firms holding decades of data aren't going anywhere, the same point Jensen made up top. So if you're an incumbent, your moat was never the software, it's the data and the trust nobody wants to leave. If you're building new, become the operating environment, or bring AI to where the data already lives. The real question isn't whether a customer can build version one without you. It's whether they want to hold the filing cabinet when version two breaks.

Salesforce CEO, Marc Benioff on the All-in Podcast

On the All-In Podcast, the hosts sit down with the SaaS GOAT Marc Benioff, CEO of Salesforce and a 40-year veteran of enterprise software. Everyone right now is debating whether AI is triggering a SaaSpocalypse. Benioff's response is that it isn’t his first rodeo. The more interesting line of questioning from the Besties isn't whether software dies, it's whether the AI hype ever shows up in the revenue, because so far the market has re-rated hard on a story the numbers haven't told. Salesforce is down 37% and its compadres are down too, yet by his account they all just posted strong quarters while trading at two times sales. His take is that there's a hypnosis around AI the financials have yet to confirm, and in the meantime a lot of software is still being sold.

In a moment of lucidity, Chamath splits the market in two: the low end is finished, but the high end, the large incumbents with deep C-suite relationships and twenty years of trust, is safe and arguably positioned to win. Benioff explains why. AI is probabilistic, so it only works when it's grounded in real data and a single source of truth. As he puts it, none of this stuff works if you don't have context. The companies that own that context, the decades of customer data locked inside their platforms (here’s the data hypothesis again), hold the thing a generic agent cannot conjure. It's the same conclusion the rest of the issue keeps reaching. The moat was never the software, it's the data underneath it.

He closes on something that is sure to land with anyone building right now. You can't get drunk on the stock price, he tells his team, because it's the one thing you can't control. The anchor is customer success, revenue, cash flow, the work itself (amen). To the founders nursing private valuations and waiting for the market to love them again, his advice is blunter: valuation is fantasyland until someone actually pays you. Focus on the customer, and let the multiple sort itself out. Coming from a man on at least his third SaaSpocalypse tour of duty, buying his own stock back while everyone else heads for the exits, the calm is hard to argue with. He has watched this story play out before, and software was still standing at the end of it.

Want to discuss more?

You could be featured in Innovation Download!

If you have an opinion or comments based on this month's topic, reply to this email with your take and we might publish your comments to our base of over 3,000 readers next month!