AI, SaaS and Crypto in 2026: Bubble, Reset or Reality Check?

AI, SaaS, crypto and the moment capital runs out of patience

In every bubble story there is a moment when everyone in the room quietly realises the trick is over. The magician keeps talking, the music keeps playing, but the audience has already seen the wires.

For AI, SaaS and crypto, that moment is arriving faster than founders and fund managers want to admit. In 2023 and 2024, capital bought the story that generative AI would unlock trillions in productivity, that subscription software would compound forever, and that blockchains would finally grow into their market caps. In 2025, warning lights began to flash across public markets, pension funds, and central banks. By 2026, the question will not be whether these technologies work, but whether their business models can carry the weight that markets and venture capital have placed on them.

This is not a “tech is dead” piece. It is a cash-flow question. When interest rates are no longer near zero, you cannot fund narratives forever; you must fund earnings.

VaaSBlock exists for people who do not have the luxury of burning money to “see what happens” next cycle. Our work on RMA™ scoring, failed chains and real-world partnerships all point to the same conclusion: trust and governance are not side quests. They are the main event.

So let us follow the money instead of the marketing, and ask a simple question.

What if the emperor’s new code is clever, but the kingdom’s balance sheet no longer supports the costume?

The AI bill is arriving, and the ROI is not evenly distributed

Across boardrooms, “AI strategy” has become a line item on every slide deck, yet the financial outcomes look wildly uneven.

Enterprise spending on generative AI has exploded. Industry estimates suggest that large organisations took their gen-AI spend from a few billion dollars in 2023 to well over ten billion in 2024, a step change in a single year. Consulting surveys show thousands of leaders across multiple industries now in active pilots or deployments, not just experiments. On paper, the story looks magnificent. Spending is up, confidence is high, and many “advanced initiatives” appear to be meeting or exceeding ROI expectations.

Look one layer down, and the picture changes.

1.1 The pilot graveyard

Analysts who track implementation rather than intent see a very different reality. Forecasts from firms like Gartner suggest that at least 30 percent of generative AI projects will be abandoned after proof of concept by the end of 2025, often because of bad data, poor risk controls, escalating costs or unclear business value. Surveys of hundreds of data leaders report that a large majority of organisations are stuck in pilot mode and cannot move gen-AI projects into production, while almost all say they struggle to show business value in hard numbers.

One CIO survey goes further, estimating that close to 9 in 10 AI pilots never reach production at all. The failure rate has little to do with model performance and everything to do with governance, integration and basic change management. From an investor’s perspective, this is the definition of dead capital: expensive infrastructure, consulting and headcount that never converts into recurring productivity or margin.

This is where the conversation picks up from VaaSBlock’s earlier piece, “AI: Threat or Opportunity? What 300 M Jobs at Risk Means for You”. That article explored how generative AI might reshape labour markets, asking whether automation would create a “useless class” or a new wave of opportunity. Two years later, the more urgent question for boards is simpler.

Not “will AI take jobs,” but “will AI justify the jobs, data centres and capex we are throwing at it.”

1.2 The optimism gap

Topline surveys tell a reassuring story. Big-four consultancies highlight “significant ROI” in the most advanced initiatives, and many leaders report that their flagship AI investments are already generating positive returns. Read only the executive summaries, and it sounds as if the AI business case is largely solved.

Drill into the footnotes, and a different pattern appears. Large enterprises admit that estimating and demonstrating business value is the single biggest hurdle for AI programmes. Data leaders talk about being “stuck in pilot” and about stakeholder fatigue as early excitement fades and costs keep rising. Even in optimistic surveys, the gap between the top decile of performers and the median is enormous.

Both narratives can be true. A small set of well-run companies is capturing disproportionate value. Many more are burning money on experiments that will never scale or never clear a real hurdle rate.

From a portfolio view, that is a problem. Equity markets have priced AI as if the median enterprise will see transformative productivity gains. The data says only a minority are executing well enough to earn those gains.

1.3 When even the vendor has to force adoption

Even inside the companies selling AI, culture does not always cooperate with strategy. In 2025, reports emerged that Microsoft was asking managers to include internal AI tool usage in performance reviews. Employees were told that using tools like GitHub Copilot and internal AI assistants was no longer optional. Around the same time, an advertising watchdog in the United States pushed Microsoft to tone down claims about Copilot’s productivity impact, noting that the cited survey captured perceived productivity rather than audited metrics.

These are not stories of a technology that effortlessly sells itself. They are stories of a vendor that must push its own people to use its flagship AI, while regulators remind it that “I feel more productive” is not the same as hard evidence.

If employees at one of the world’s biggest AI software companies need that sort of nudge, it is not hard to imagine the friction inside a regional bank, a manufacturer, or a government department with less technical comfort and more regulatory risk.

Open source is compressing the AI margin story

Markets have behaved as if frontier AI will behave like a natural monopoly, with a small handful of model providers capturing most of the profits. The last two years of open source development tell a different story.

2.1 Open models, lower costs

Open source has always been an economic pressure valve. Studies from Linux Foundation Research and others show that organisations relying on open source software spend significantly less on technology while still increasing productivity and innovation. In AI, the pattern is repeating. Surveys of hundreds of organisations find that almost 9 in 10 adopters are using open source AI in some form, and a clear majority say open models are cheaper to deploy than proprietary ones. Many adopted them specifically to reduce costs or avoid lock-in.

Benchmarking platforms tracking model performance now show community and challenger models approaching or matching the quality of closed systems on many workloads, especially once fine-tuned on domain-specific data. For cost-sensitive enterprises, particularly outside the United States, “good enough” open models are becoming hard to ignore.

2.2 International pressure on US margins

The threat to US model economics is not only open source in the abstract. It is the emergence of aggressive international challengers. Developers in China and elsewhere have released open-weight models that are good enough for many enterprise tasks and undercut US pricing. Download data shows a surge in adoption of such models, with usage shifting away from US-centric, frontier-only stacks toward lower-cost, regionally optimised systems.

This compression matters because so much of the Big Tech valuation premium is tied to AI margin assumptions. When investors model the future earnings of hyperscalers, they do not just look at cloud revenue. They look at high-margin AI services that ride on top. If open models take a growing share of workloads, that future margin stack looks less secure.

2.3 The myth of the model moat

Inside the AI gold rush, it is fashionable to say that “the model is the moat.” In practice, the moat is shifting toward data, distribution, and the unglamorous work of change management. Models can be downloaded, replicated and improved by competitors. Contracts with large customers and the ability to change how thousands of people work are harder to copy.

This brings us back to a recurring VaaSBlock theme. Governance and trust matter as much as technology. In the earlier piece on AI and jobs, the focus was labour. Here, the focus is financial. If you are betting on an AI vendor, the question is not only “how advanced is the model,” but “how credible is the business in turning that model into durable, profitable usage.”

In 2026, that distinction will decide which AI companies look like utilities with thin margins, and which still earn a genuine premium.

SaaS discovers gravity

For more than a decade, software as a service felt like a law of nature. Subscriptions would rise forever. Net dollar retention would stay comfortably over 100 percent. Exit options would always be available, either via strategic acquisition or a friendly IPO window.

The last three years have dismantled that story.

3.1 Growth slows, funding freezes, exits vanish

Bankers who track software IPOs point out that only a small handful of SaaS companies have managed to go public in the last three years, the weakest environment for new software listings in decades. Of the wave of COVID-era listings, most now trade below their IPO price or have been taken private. Late-stage private valuations, meanwhile, often remain higher than what public markets are willing to pay, which closes one of the main exit doors.

On the private side, multiple analyses show that SaaS and enterprise software funding has fallen sharply from its 2021 to 2022 peak. Investors who once treated recurring revenue as an automatic premium are now more interested in burn multiple, gross margin and path to profitability. Cloud and SaaS indices still attract capital, but the tone has shifted from “land grab” to “show me the cash flow.”

BetterCloud’s State of SaaSOps report captures this shift from the customer side. For the first time, the number of SaaS applications in use at companies declined rather than grew. The loudest directive to IT is simple, cut costs and consolidate tools.

3.2 The 18 million dollar waste problem

When budgets tighten, CFOs start looking at every line item. Zylo’s 2024 SaaS Management Index, based on tens of millions of licenses and tens of billions of dollars of spend, finds that enterprises waste an average of 18 million dollars per year on unused SaaS licenses. On average, companies are using roughly half the licenses they pay for.

If you are a founder building a 300 dollar per user per month tool, that number should terrify you. It means enterprise buyers are no longer judging SaaS contracts as “the cost of doing business,” but as discretionary spend that can be cut when something cheaper or more flexible appears.

3.3 AI makes subscription bloat harder to defend

This is where AI loops back into the SaaS story. The same BetterCloud and CIO surveys that talk about consolidation also highlight the role of automation. Large language models can now handle tasks that used to require dedicated subscriptions. Report drafting, basic analytics, summarisation and even some project management coordination can be stitched together with a general-purpose AI and lightweight integration.

This does not kill SaaS, but it moves the bar. To justify its price, a specialist tool must deliver much more than a generic LLM can. It must plug deeply into workflows, offer defensible data, or provide compliance and controls that general AI cannot match.

In 2025, public markets started to reflect this reality in compressed revenue multiples and lower expectations for growth. In 2026, that discipline will show up in boardrooms as failed rounds, flat rounds and structured terms for companies whose products can be partially replaced by AI and whose unit economics never made sense without cheap money.

Crypto: when trust and treasury collide

Crypto has already lived through several “emperor has no clothes” moments. Each time, the market crashed, then recovered stronger in narrative terms, if not in fundamentals. The current cycle is different because the participants have changed. This is now a story of ETFs, pension funds and listed companies, not only retail traders and offshore exchanges.

4.1 Volatility and the illusion of digital gold

In late 2025, Bitcoin fell by more than 30 percent from its recent peak in a matter of weeks. Spot Bitcoin ETFs saw record net outflows as institutional holders headed for the door, and the broader crypto market lost hundreds of billions of dollars in paper value. For traders, this is familiar. For treasurers and CFOs who tried to treat Bitcoin as a long-term corporate reserve asset, it is a different experience.

The promise of “digital gold” was that Bitcoin could serve as a relatively stable store of value over long horizons. In practice, balance sheets that leaned too heavily on it became gearboxes for volatility. Mark-to-market swings of tens of percent created headaches for auditors, risk committees and regulators, particularly in listed companies.

4.2 Digital asset treasury firms unravel

In the last few years, a cluster of listed companies adopted what analysts call a digital asset treasury strategy. They raised equity or debt, used the proceeds to buy crypto, and marketed themselves as leveraged ways to gain exposure to Bitcoin.

Data compiled by market researchers and reported by mainstream financial press shows that these firms reached a combined market cap in the hundreds of billions of dollars at their peak before falling dramatically as crypto sold off. Some of these companies have already started dumping tokens to fund buybacks or service debt. In one striking case, a semiconductor manufacturer sold more Bitcoin than its own market capitalisation, a situation where the treasury asset had effectively become larger than the operating business.

On-chain analysts warn that many digital asset treasury firms could go under if prices stay depressed, because the market premium they once enjoyed has vanished. None of this means Bitcoin is going away. It does mean that treating a highly volatile asset as the foundation of a listed company’s capital structure is a governance risk, not a clever hack.

4.3 Code is not a company

This lesson extends beyond corporate treasuries into the world of blockchain projects themselves. In “Code Isn’t a Company: Kadena’s Shutdown and the Cost to Trust” VaaSBlock dissected how a technically capable blockchain could still fail investors and users when governance, communication and business model fundamentals were neglected. You can have an elegant protocol and still destroy value if you do not manage disclosures, expectations and risk in a way that traditional capital understands.

The current crypto cycle has produced many similar cases, from exchanges that rebranded leverage as “yield” to projects that treated token emissions as revenue. As interest rates normalise, those shortcuts no longer work. Investors and regulators are asking basic questions about cash flows, reserves and the rights attached to each instrument.

4.4 Where crypto actually makes sense

This is where use cases that look boring on social media start to matter. One example is the partnership between VaaSBlock and Swift Cargo, a global relocation provider. In that collaboration, outlined in “Swift Cargo joins forces with VaaSBlock to launch Crypto Payment Solutions for Global Relocation Services”, blockchain is not treated as a speculative asset. It is used to improve how international relocation invoices are paid, reconciled and verified.

For relocation firms and distributed teams, being able to settle cross-border moves with digital assets and on-chain proof of payment provides clear operational value, such as faster settlement, fewer errors and better audit trails. Readers who want to see how this looks in practice can explore Swift Cargo’s own site for a commercial view of crypto-enabled relocation services.

In other words, the future of blockchain is not in “number go up,” it is in reducing friction where traditional financial systems are slow or fragmented. In 2026, capital will keep flowing to that kind of use case. The projects that relied solely on story will find the next round harder to raise.

The macro backdrop: when everyone is worried about the same seven stocks

Markets do not reprice technology stories in a vacuum. They reprice them when concentration and leverage begin to bother the people who have to manage systemic risk.

5.1 Central banks and the AI bubble question

Recent financial stability reviews from the European Central Bank and other institutions have warned that stretched valuations and high risk concentration in equity markets could make them vulnerable to sharp corrections, especially when a small group of tech stocks drives a disproportionate share of returns. In more recent remarks, the ECB has singled out US tech and AI names, noting that valuations look stretched and that investor fear of missing out is pushing prices higher without a comparable improvement in underlying certainty.

The Bank of England’s Financial Policy Committee has echoed these concerns, stating that global markets could face a sharp correction if expectations about AI are revised downward. The International Monetary Fund has made similar noises about concentration and speculative excess in AI beneficiaries.

The UK’s Office for Budget Responsibility has gone one step further. It modelled what would happen to public finances if an AI-fuelled tech bubble burst and concluded that such a crash could create a multi-billion pound borrowing shock, putting government fiscal targets at risk. These are not blog posts from contrarian hedge funds. They are warnings from the institutions paid to care about financial stability.

5.2 Pension funds quietly leave the party

If you want a practical indicator of how seriously those warnings are taken, look at pension funds. Reporting from mainstream financial media shows that several large UK pension schemes, managing hundreds of billions of pounds, are cutting exposure to US equities, especially AI-driven “Magnificent Seven” names such as Nvidia, Alphabet and Meta. The rationale is straightforward. Valuations look stretched, concentration risk is high, and retirees cannot afford a 30 to 40 percent drawdown if the AI story disappoints.

Funds are reallocating toward more diversified geographies, defensive assets and private markets, while those who stay in US tech are explicitly layering on downside protection. When long-horizon investors who are supposed to ride out volatility start trimming exposure to the very names that have powered index returns, it is a sign that the story has run ahead of what cautious capital is comfortable with.

5.3 The Magnificent Seven as future utilities

Some analysts already argue that the largest AI beneficiaries will slowly morph into something closer to utilities. The logic is that massive capital expenditure on data centres, chips and infrastructure will eventually make them look like old-world network businesses, with stable but lower returns on incremental investment.

If that happens, the multiple that markets are willing to pay for those earnings will compress. Investors will not pay a hyper-growth valuation for a utility profile, no matter how clever the AI marketing around it.

In that environment, even a healthy AI business can see its stock price fall, simply because the future is less explosive than the slide deck once implied.

2026: when capital runs out of rope

Put the pieces together, and 2026 looks less like a catastrophic crash, more like a forced honesty session.

• AI projects that have not found a path to production or measurable value will be culled.
• SaaS products whose unit economics only worked in a world of infinite budgets will discover how fragile their net retention really is.
• Crypto projects and treasury strategies that relied on reflexive story-driven flows will be stress tested by volatility and by regulators.
• Public markets will adjust the multiples they assign to even successful tech names once the growth story shifts from unlimited to simply “good.”

For builders and early-stage investors, that sounds bleak. It does not have to be.

6.1 What survives a reality check

The projects that come through this cycle stronger will share a few traits.

They can show their work. Not in a pitch deck, but in their numbers. They can point to reduced headcount per unit of output, faster cycle times, higher customer retention, or margin expansion that would not exist without the technology. They can do it in language a CFO and a regulator understand.

They treat governance as a feature, not a checkbox. This has been a recurring theme in VaaSBlock’s work, including the RMA™ scoring model and deep dives into failed chains. Modern investors will not fund a black box for long. They expect clear disclosures, sensible token or equity structures, and a realistic view of risk.

They respect the difference between code and company. The Kadena case study is a warning that engineering quality does not automatically translate into a resilient business. In AI and SaaS, the same applies. A brilliant model or product can still fail if customer acquisition costs are unsustainable or if switching costs are low.

They earn their place in a stack that now includes general AI. If an LLM can replicate 60 percent of what your tool does with a bit of configuration, your pricing and product strategy must acknowledge that reality. The remaining 40 percent has to be indispensable.

They build for real-world constraints. High energy costs, regulatory uncertainty, data residency rules and ordinary user scepticism. The Swift Cargo example shows how crypto can be deployed in a tightly scoped way that respects those constraints and still creates value.

6.2 A simple checklist for founders and investors

If you are running a company or thinking about backing one in AI, SaaS or crypto, the following questions are uncomfortable but useful:

• Unit economics: Can this product still be profitable if customers cut seats by 30 percent or demand a 20 percent price reduction once their CFO reviews AI alternatives?
• Adoption reality: Are employees using the tool because it helps them, or because management is forcing usage into the performance review process?
• Open source pressure: What happens to your pricing power when open models or open alternatives close half the gap in quality again?
• Governance and disclosure: Would your capital structure, treasury decisions and reporting survive a forensic review from a sceptical regulator or a conservative pension fund?
• Exit dependency: Is your business viable as a cash-flow engine, or does it only make sense in a world where you can hand the problem to public markets through an IPO at a high revenue multiple?

If too many answers rely on “the market will keep rewarding growth” rather than “the business makes money,” you are not building for 2026. You are building for 2021.

From AI hype cycle to trust: governance that survives the 2026 reality check

In earlier work, VaaSBlock framed AI as a double-edged sword for workers, a force that could both displace and empower. The argument was that the real challenge was distribution — who benefits from AI-driven efficiency gains and who is left behind.

The same is now true for capital in AI, SaaS and crypto. These technologies have already delivered immense value and will continue to reshape industries and create new fortunes. None of that is in doubt. What is in doubt is whether the current crop of AI-, SaaS- and crypto-linked stories and valuations has kept pace with the slower, practical work of building resilient, cash-flow-generating businesses.

In that sense, 2026 is not the year the AI, SaaS and crypto markets are exposed as a fraud. It is the year the court begins to insist on basics again: cash flow, governance, accountability and real customers who would miss the product if it disappeared.

For VaaSBlock, this is not a new theme. It is the reason the RMA™ blockchain credibility framework exists, the reason we dissect failures like Kadena, and the reason we highlight partnerships like Swift Cargo that turn crypto into infrastructure instead of theatre. The next cycle will reward organisations that can pass a serious governance and revenue test, not just a hype test.

If the last cycle was about who could tell the most compelling AI or crypto story, the next will be about who can keep telling the truth about risk, governance and returns when the music stops. That is a much healthier game to play for investors, founders and regulators alike.