The AI earthquake continues to rumble across the globe, creating opportunity, disruption, hope, and fear in its wake. The enormity of this technology revolution dwarfs all others. During the go-go years of the internet bubble, VCs invested $80B, compared to today's AI VC warriors, who have invested 6x that at $476B. AI has created 39,100 newly funded startup companies, and there is no slowdown in sight. Analysts project that AI CapEx could hit $1.25T annually by 2031, which would require roughly $6T in annual AI revenues for normalized, profitable financial outcomes. That equates to the $6T generated in '25 revenues by the 13 largest companies in the world – including Amazon, Walmart, Apple, and Saudi Aramco – that have been around for 50 years.

The ten largest AI companies, including Anthropic, OpenAI and SpaceX, have grown AI revenues from nothing a few years ago to a current run-rate of $163B. Muse, Meta's new consumer knockoff of Claude Cowork, reached 900,000+ downloads in its first week. Goldman estimates there will be $7.6T invested in AI infrastructure over the next six years. 70% of the 100 largest stocks' market value is now concentrated in AI, versus 36% in tech at the height of the internet bubble in 2000. Every business, government, and worker across the world is now grappling with AI at some level. While exploiting the benefits of AI has real challenges and will take much longer than the pundits predict, it will eventually bring massive productivity improvements for businesses, workers, and consumers globally across software development, healthcare, manufacturing, defense, service industries, and simple everyday chores – for decades to come.

A battle is currently playing out over who will deliver the real value and reap the lion's share of profits from delivering on AI's promises. Over the last year, you were told OpenAI, Anthropic, and xAI would be the champions. Maybe, maybe not. Frontier lab market share has cratered in the past two years, with closed-weight models falling from 78% to 28% of OpenRouter token volume and DeepSeek alone accounting for more than 60% of Vercel token usage in the past two weeks. NVIDIA, which has been the successful arms dealer selling their chips to all, has now moved up the architectural stack – from chips to open LLMs – with its acquisition of Hugging Face. Enterprise AI customers are reaping the benefits of this intensely competitive battle, which is rapidly improving machine intelligence, driving down costs, increasing LLM options, and essentially commoditizing the vast majority of LLM chores.

At the same time, specialization, localization, and distillation are producing smaller, purpose-built models that can absorb much of the frontier's capability and handle specific tasks at a fraction of the cost. Orchestration layers then route each task to the lowest-cost model that can do the job, saving expensive frontier models for only the hardest problems and driving the cost of intelligence down even further. The result will be a more democratic and capitalistic AI market, where customers choose the best model for each task and both SaaS companies and enterprise customers retain sovereignty of their workflows, data, and customer relationships.

The magnitude of the potential for AI in the future is undeniable. That said, we have overplayed, overinvested, and overstated the near-term benefits and financial payback of what is realistically possible for AI to deliver. Business leaders across the globe and across industries have embraced AI. 63% say it is fully deployed, but only 21% are actually getting a return on their investment.

At AGC Partners, the team trains on Claude every week and uses it for much of our work, making us more productive and improving our product and service, but it hasn't translated into materially lower labor costs. In our highly competitive industry, as in many industries, you have to up your game because of AI and so do the elite players around you. Therefore, AI does not miraculously increase revenues or reduce cost. Companies can show very impressive productivity gains in individual tasks without creating more value for the customer. Coding can get faster, research can get faster, and pieces of a workflow can improve without materially changing the overall process or customer outcome. The operating model has to evolve so those productivity gains actually flow through to the customer and show up as revenue, margins, and customer value.

Many of the Fortune 100 have jumped headfirst into the AI tractor beam, but many are now slowing – if not fully reassessing – their comprehensive AI strategies and considering completely new approaches to try to capture this elusive ROI payback for all the money spent. The gold rush to build data centers has sucked up trillions and trillions of dollars in equities and debt that have supercharged global equity valuations. But the buildout is now running into local opposition and tougher economics as LLMs and tokens commoditize and the revenue and cash flow assumptions behind 20-year projects get harder to defend.

The top 100 US stocks are valued at $53T, with $23T of that value having been created by AI stocks in just the last three years. 70% of that stock market value is from AI-engaged companies. Those 43 AI companies are trading at 36x earnings, even with those earnings at historic highs due to the gold rush infrastructure buildout that may very well not play out as planned. AI is lifting the U.S. and global economy singlehandedly. How large will the blast zone be if the AI promises come up short?

On the good news front, the prediction that frontier models and AI startups would decimate SaaS has obviously not come true. Our SaaS index was worth $4T in September '25, fell to $2.6T at the height of the SaaSpocalypse in April '26, and recently made it back to $4T. The SaaS incumbents have continued to beat projections, accelerate growth, retain strong retention metrics, and prove far more resilient than feared.

Don't get me wrong, there are many young AI companies crushing it – growing from zero to tens or hundreds of millions of dollars of revenue in just a few years. We saw this firsthand with our client Workstreet. These AI-native innovators are not only taking share from industry incumbents but also expanding the TAM for everyone. Harvey has grown annualized revenue from $190M to more than $350M since January, and Sierra went from $100M to more than $150M of ARR in less than three months. Cognition, founded less than three years ago, just crossed $1B in annualized revenue after raising more than $2B at a $48B valuation. That success comes with enormous expectations, and these companies will need to sustain extraordinary growth to grow into today's valuations.

The incumbent software companies' natural moats shown below are in most instances far broader and more effective than the SaaSpocalypse evangelizers fully understood. This is not to say the battle is over, but it is to say it is not a foregone conclusion that the LLMs and the startups crush the SaaS incumbents. It is much more likely that it will be a blended world where the most forward-leaning management teams continue to lead in their sectors with both in-house and acquired AI capabilities, and those that hesitate will be outplayed by the AI startups and other incumbents leading the charge.

SaaS companies still live under the cloud of AI fear and slower growth. One is a future boogeyman that you can't prove or disprove, and the other is basic math. If you are growing half as fast as you used to grow, you are worth a lot less. Public SaaS multiples for AGC's 100-company index below $10B are trading at 2.7x revenue and 13x EBITDA. This has put almost a full stop on SaaS IPOs with only three YTD, the most recent being Bending Spoons on July 1. The combination of slower growth, AI uncertainty, and sellers' valuations still anchored to prior highs has also shut down the PE take-private market: the last $1B+ SaaS deal was OneStream, closed in April, versus the 12+ take-privates a year that we have seen since 2020. Recently, growth has become a better barometer of valuation than Rule of 40 – partly because growth is so hard to deliver these days. That said, the days of financings or acquisitions of mega loss-makers or inefficient growth in traditional SaaS are long gone. Simply put, the spread in valuations for a SaaS or AI company is anywhere from 1-20x revenue as the gap between the haves and have-nots has widened. A company with no growth, meager profits, light tech, and limited moats could be a 1-2x play on the customer base and the future cash flows, whereas you might get 12x revenue on a 60% growth, breakeven rocket ship.

SaaS incumbents, both public and private, are nevertheless in the midst of retooling their offerings to be AI leaders. That process is still in the early innings – it will take quarters, if not years, to make this initial monumental transformation. Simultaneously, the software world needs to grapple with the fact that many of their markets have become so saturated that growth is down from 25%+ to 11%. That, along with the AI transformation, has reduced valuations and effectively thrown the tech PE world into M&A paralysis. If your portfolio company either has i) not completed the AI transformation, ii) not put up strong Ro40 and gross retention metrics, or iii) is still marked at COVID-era high valuations, then that portfolio company is not in a position to sell. Even if it is good on those three markers, you may not want to sell because it is one of the best assets in your portfolio.

The global tech PE world now has roughly 4,600 majority-controlled companies within their portfolios held by roughly 250 funds. Those funds year to date have only exited 113 companies and made 138 new platform investments. Those annualized numbers are down roughly 26% on exits and 47% on platform investments. More importantly, you are looking at a backlog of 30 years to sell those 4,600 majority-controlled companies with the current velocity of sales. That obviously doesn't work for 10-year funds with a 6-year average holding period. So back to the good news, there will be a pickup – and a very large pickup at that – as valuations get marked to market, AI transformation happens, and market participants capitulate to a new world order of lower growth, lower returns, and lower values.

With the paralysis of the tech PEs in the AI & SaaS M&A market, the strategics have increased their market share of M&A dramatically. In '25, strategics did 28% of the acquisitions as compared to 63% in '26, while the PEs have fallen from 72% to 37%. Excluding the Cursor aberration, overall tech M&A value is down to a $219B run-rate from $313B in '25. Good thing the strategics are jumping back into the game.

For those companies that have an AI story with strong metrics, the market has never been this good. The premium 2026 strategic acquisitions of AI plays are 10x+ revenue with a median of 17x. AGC just sold two in that range. PE funds with their record $555B in dry powder are paying up for exceptional-growth AI plays with strong Ro40 metrics at a median multiple of 12x. AGC recently announced the sale of Workstreet, which was formed three years ago, and generated great interest, an exceptional valuation, and a rapid process. There are not enough companies going to market these days, but when you have a hot one with strong AI, fast growth, and profits, there is more demand than we have ever seen.

AGC tracks 66,000 global AI and SaaS companies of scale across the world. We closed 28 AI and SaaS deals last year, and we have 18 deals closed or under LOI to date this year. All of these companies are facing the same unanswerable question of how AI is going to affect their markets and company 2-5 years down the road. We do our best to predict the unpredictable with all the intelligent backup we can muster, but the reality is, after a year of the AI earthquake, the most forward-leaning SaaS incumbents continue to hold their position and outperform. That – along with a deep and articulate understanding of the current dynamics in AI and a strong roadmap forward – is the most important proof point. If you are not aware of that small AI startup that is kicking butt in your space, that is a problem. If you have the best agentic solution in your space, albeit a tiny portion of your revenues, that is impressive. Most of our clients are well on their way on this roadmap. In many instances it makes sense to stick to your knitting and just build your company organically, but in other situations where the market has been saturated and growth is slowing, it could well be a perfect time to make consolidating acquisitions of competitors in your space. This is an opportune time for those confident in their understanding of this very complicated industry to make bets on their convictions.

AGC is currently brainstorming with founders, CEOs, and GPs on how and when to optimize the path forward, whether it be no action today or full speed ahead on a deal. Our Partners' deep AI and subsector knowledge, highly active live deal experience, and strong strategic and PE relationships empower our Partners to be exceptional advisors for CEOs trying to chart a course through these challenging waters. Whether you're a SaaS incumbent or a fast-growing AI entrant in Cyber, GRC, or any of the 50 verticals we cover, the AGC advisory team is deep in knowledge and relationships, ready to go 24/7, and able to power through those bumps in the road to bring about great outcomes.