Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Peter Day, General Partner, super{set}
If 2025 was the year that AI captured the popular imagination and moved to the center of the tech sector discourse, 2026 will be the year where expectations will need to be met. We’re in exciting times where the pace of tech evolution is super fast paced. To date, the AI charge has been led by major innovators including OpenAI, Anthropic and Nvidia, with plenty of healthy competition from legacy tech giants like Google and Microsoft. Sectors and disciplines like marketing, software development and financial services are already being transformed by AI.
At the same time, news reports through the year have documented sobering setbacks: an MIT study found that 95% of of generative AI pilots at companies are failing to turn a profit; in November, Microsoft financial reports reflected that its investment in OpenAI resulted in a net loss of $11.5 billion or more for the quarter. Despite the turbulence, the world now knows what AI is and corporations are starting to invest and experiment much more to prove out the value. Startups are getting a lot of attention and garnering deeper investments out the gate. Products are entering markets at a lightning fast pace. What will 2026 have in store for the AI wave? Here’s a few thoughts and some predictions.
PREDICTION #1: Commodification & Consolidation
Foundation model companies are increasingly becoming commodified, prompting many AI companies to move up the stack into agent-building solutions for very specific vertical problems to maintain differentiation. Initially, companies tried building their own integrations for features like tool calling, but now standardized protocols are emerging-threatening most horizontal infrastructure plays.
The challenge is that when tech giants like OpenAI, Google, and Microsoft can give away similar capabilities for free, competing becomes extremely difficult. These players have deep pockets and established distribution channels-the real moat comes from user adoption, which creates a testing and learning flywheel and potential data moat. Competing with giants who control distribution makes the horizontal bet increasingly tough-for these reasons, many startups are pivoting from horizontal infrastructure plays to use-case-specific approaches.
PREDICTION #2: Avoiding Disintermediation as Consumer AI Adoption Accelerates
The next couple of years will see a battle in the consumer AI space similar to the streaming wars. When Samsung TV first launched, users could search across all streaming platforms regardless of which app hosted the content-Hulu, Disney+, or Netflix. It was a great consumer experience until content providers fought back, forcing users to stay within individual app environments to maintain their attention and avoid disintermediation.
We’re likely to see this dynamic replay with agentic AI apps, creating numerous frenemy relationships. Consider Booking.com in the travel space: they want bookings from people using ChatGPT, but they don’t want to become a commodity provider losing direct customer relationships. If ChatGPT becomes the front door to consumer goods and services, everyone will want to be included while simultaneously protecting their direct customer access. This tension will intensify because monetizing AI is challenging. The technology is expensive, yet consumers won’t pay much more than a $20 monthly subscription. Startups will need indirect monetization through ads or commissions, driving significant strategic decisions going forward.
PREDICTION #3: Messy Alignment Between AI Adoption and Infrastructure Build-Out
AI startups face a fundamental challenge: customer adoption and infrastructure requirements rarely correlate neatly. While achieving adoption is always hard, AI faces unique complications beyond customer willingness to pay-the backend infrastructure requires massive capital investments in data centers and foundation models.
Over the long term, these infrastructure investments make sense because humanity will inevitably demand more compute resources, just as we demanded more bandwidth in the early commercial internet days. However, the timing creates turbulence. If compute becomes scarce and expensive just as your AI startup scales, that’s a problem. Conversely, building capabilities ahead of adoption makes it harder to justify those investments while struggling to gain market traction. From a stock market perspective, these infrastructure investments may look foolish in the short term-imagine investing $50 billion in data centers when customers aren’t willing to pay for it. But over longer time horizons, the investments will prove necessary.
The 1990s offer a cautionary parallel. Companies laying cables for increased bandwidth made huge infrastructure investments, then got caught in the dotcom bust-Nortel’s bankruptcy being the poster child. Of course, we now rely on all that infrastructure. The question is: who’ll be left holding the bag this time? Keep in mind that the shelf life of the GPUs that power AI systems is considerably shorter than the optical fiber supporting today’s digital networks.
PREDICTION #4: Incremental Progress Toward Agentic Offerings & New Compute Devices
Many enterprise AI projects will likely fail to deliver expected ROI. Instead, we’ll see breakout consumer experiences that are truly agentic-taking actions rather than just providing information. Think agents that actually carry out multi-step workflows, not chatbots that merely convey information. In the business sense, we’re likely to see full stack AI companies emerging. What would that look like?
To date, we’re seeing existing business entities-marketing agencies, law firms, and financial services companies, for instance-incorporating AI into established work processes. But what if instead of AI developers selling agentic capabilities into existing companies, startups develop their own agentic capabilities to compete directly with legacy players. In this way, the startups can avoid the “change management tax” and grow unencumbered by providing better, faster, cheaper services powered by AI in under-disrupted sectors. Getting to this point is likely to be several years out, but we’re already seeing moves in these directions.
As for devices, consider a bread-making example: when my dough felt dry, I showed ChatGPT a video and it advised adding olive oil, then confirmed when I’d added enough. The bread came out great. But the mobile phone form factor is terrible-pulling out your phone with flour-covered hands is impractical. The ideal compute hardware needs to see what you see and function without requiring screen navigation. Perhaps voice-activated, like a Star Trek communicator badge. It must be listening, feel natural, and be super fast. The best example today is probably Meta AI glasses, which combine an integrated camera, microphone array, and open-ear speakers, integrated with a smartphone app. The competition for the next compute platform will intensify in 2026, and reach fever pitch over the next 5 years.
PREDICTION #5: Agentic AI and the Democratization of Luxury
Agentic consumer experiences will be where AI’s power first manifests for general consumers, likely in obvious applications like travel arrangements. AI excels at solving multistep processes that humans find semi-unpleasant or tedious. Current AI travel tools aren’t quite there yet. They get you to the irritating part-recommending a restaurant but still requiring you to navigate their reservation system and see if there is availability. What’s needed is a simple yes-or-no interaction.
In my former corporate role, I was fortunate to have an incredible executive assistant who handled everything. He’d say, “you’re in Manhattan for meetings March 3rd-4th. Here are three hotels at our corporate rate and five restaurant options.” Just yes-or-no decisions with multiple choices, and he’d handle all transactions using the expense account credit card. When agents can replicate this interaction, everyone will access the concierge services previously available only to executives and wealthy individuals. This democratization of luxury is what should happen-and I believe it will.
What’s Next?
How and whether any of these predictions will play out remains to be seen, but we’ll know soon enough. What’s not in dispute is that AI is already making a massive impact across sectors and markets, and will continue to do so for years to come.
##
ABOUT THE AUTHOR
Passionate technologist and coder, experienced engineering and product leader, machine learning PhD, and occasional speaker. Formerly CPTO at Quantcast.
