What We’re Watching: 645 Ventures’ Inflections and Predictions for 2026

2026 is shaping up to be a defining year for technology startups, driven by the rise of AI and multiple other inflections that are catalyzing the formation of transformational startups. At 645, we apply a research-oriented approach that enables us to understand key inflections and be a prepared mind when we engage with startup founders.

In this piece, we share the inflections we’re watching, where we expect to see the next wave of category-defining businesses emerge, and key predictions for 2026.

If a prediction from one of our team members resonates with you as a founder, investor, tech enthusiast, we’d love to hear from you. Feel free to reach out directly and reference 2026 VC Predictions.

AI’s Bottleneck Shifts to Inference and a New Infrastructure Layer Emerges

Contributed by Aaron Holiday & Britt Binler ([email protected] & [email protected])

In 2026, the center of gravity in AI will shift more heavily from training the model to using the model: inference. Once a model is trained, inference is what happens every time someone chats with an assistant, generates an image, or runs an agent.

Because inference happens millions or even billions of times for every model that’s trained, it is becoming the dominant driver of compute cost, hardware demand, and system slowdowns. As AI becomes integrated into every consumer and enterprise workflow, the major limiting factor won’t be how advanced models are, instead it will be how efficiently we can serve them at scale.

This technology shift is creating increasing opportunities for a new generation of companies that could become as important as the hyperscalers are today. Building the inference layer from the ground up, making it super fast, cheaper, and less dependent on expensive GPUs will be a game changer.

This wave is being led by creative founders re-envisioning how to restructure models to run efficiently on CPUs or orchestrating complex multi-agent workloads across data centers. If there is going to be as much AI demand as everyone is betting on, then there will be very large inference companies making sure AI experiences are delivered as efficiently as possible. We are seeking to invest in founders building companies at the forefront of this wave.

Agentic Financial Services Infrastructure Will Unlock a New Wave of Fintech Innovation, while AI Business Models Will Face a Reckoning from Investors

Contributed by Nnamdi Okike ([email protected])

The agentic future of financial services is currently being built, and in 2026 we will see companies introducing real innovations in commerce, payments, lending and banking. The infrastructure layer for payments and financial transactions must be built first. This layer enables agent-to-agent communication throughout a financial transaction, from product discovery, to selection, to negotiation, to payment execution and closing.

New startups will build this infrastructure over the next several years, partnering with incumbents and regulators to create the rules governing agentic financial services, but also innovating around these stakeholders to unleash customer and user value. AI commerce and banking applications companies will build on top of this infrastructure, ushering in new business models and products and services that previously were not available to consumers and businesses.

Just as the previous mobile and SaaS waves enabled companies like Revolut, Nubank, Chime, and Klarna to be built, the current AI wave combined with technologies such as stablecoins will enable massive new fintech businesses to be founded and scaled over the next decade. Startups are well positioned to win in this market, because they aren’t dependent on old infrastructure and outdated business models.

In addition, by the end of 2026, the financial markets will begin to reckon with the realities of AI business models and capital will begin shifting away from the capital-intensive, “spend at all costs” approaches of hyperscalers and foundation models. Both public and private investors will regain discipline and apply a stricter lens to their AI investments.

While the correction will begin with hyperscalers and foundation model companies, it will eventually impact the entire AI stack, including application companies.

This will have a ripple effect across the tech markets. The receding tide will result in reduced valuations in 2027 in both public and private markets, and may trigger a market downturn.

Healthcare AI Will Be Won by Vertical, Compliance-Native Agent Platforms

Contributed by Jon Smith ([email protected])

Healthcare is entering a phase where data liquidity, interoperability mandates, and workflow automation pressures are converging to reshape how clinical and administrative work gets done.

As organizations confront rising labor shortages and increasingly complex regulatory requirements, they are turning toward AI-driven automation to fill critical operational gaps. This shift is creating the perfect conditions for a new generation of deeply vertical, healthcare-native agent platforms to emerge.

In healthcare, history has repeatedly shown that horizontal platforms eventually give way to vertical, healthcare-specific winners, whether it was Salesforce yielding to Veeva in life sciences CRM, Zoom falling short where Teladoc built clinical-grade virtual care, or Qualtrics unable to match Press Ganey’s regulatory-aligned patient experience stack.

In 2026, I believe we will see the same pattern repeat in AI as horizontal agent platforms struggle to meet the compliance, workflow, and data-integrity demands of healthcare. The market will require purpose-built, HIPAA-compliant agent tech stacks that understand clinical nuance and solve real provider and payer operational complexity. I am prioritizing founders building companies capitalizing on this inflection.

Domestic Stablecoin Usage Finally Hits Its Moment, and Stablecoin Payments Will Feel Like Apple Pay (or Better)

Contributed by William Hess ([email protected])

One of the biggest shifts I’m excited about in 2026 is the rise of domestic stablecoin usage - not just for cross-border transfers or bank-to-bank settlement, but for everyday spending. Financial institutions are already leaning in, with Visa projecting stablecoin supply approaching $2T and transaction volumes growing 10× over the next few years.

Yet the wild part is that in-person stablecoin payments are still basically at zero adoption today, even though merchants are getting crushed by 3%+ fees, slow settlement times, and fraud risk built into the legacy rails - so the opportunity feels wide open.

In 2026, I predict that stablecoin payments will start to feel as effortless as tapping your phone with Apple Pay or sending money on Venmo. Right now the UX is terrible - most crypto wallets rely on QR codes that take nine clicks and eleven seconds to complete a transaction, and the NFC layer is locked down by Apple, Google, Visa, and Mastercard, who have no reason to disrupt their own economics.

But we’re finally seeing crypto-native tap-to-pay rails that settle in ~100 milliseconds, remove all intermediaries, and cut acceptance costs by 100× for merchants. Once that experience reaches consumers, stablecoins stop being “crypto payments” and start being the easiest way to pay—full stop.

Generative AI Will Power the Next Wave of Content Platforms

Contributed by Yasmeena Faycurry ([email protected])

There is a lot of excitement around how GenAI will reshape both the creation and consumption of content (video, gaming, music, etc.), and lead to the emergence of generative-native platforms. As costs collapse and creativity becomes unbound by the physical world, I predict a shift from handcrafted content platforms to prompt-native, iterative platforms where the creations themselves form the network rather than the individual user accounts.

I am prioritizing founders creating platforms purpose-built for generative content at scale. As GenAI video reaches true parity with Hollywood-grade production (becoming dramatically more reliable, expressive, and controllable), AI-generated content will move from a “slop-y” novelty on TikTok/Instagram to something people consume with genuine delight, embracing its unique capabilities unconstrained by the economics and physics of the real world.

This shift will lead to the emergence of new platforms dedicated to generative-native storytelling, with the first breakout players likely scaling by 2027.

Challengers Will Make Meaningful Strides in AI Chip Design, and Jailbroken AI coding models will power multiple critical cyberattacks in 2026

Contributed by Vardan Gattani ([email protected])

In 2026, Nvidia’s AI hardware-software moat will continue to proliferate, but challengers will make meaningful strides in AI chip design. In the U.S., current mega Nvidia customers will continue trying to break their dependence on the giant with their own specialty hardware, similar to Apple. These players will mostly continue to invest in purpose-built ASICs like Google’s TPUs and Amazon’s Trainium/Inferentia products to advance the economics and performance of their own AI training/inference.

If a breakthrough occurs in 2026, Meta, Microsoft and OpenAI will be emboldened to accelerate their own projects. Globally, China will see the most progress competing with Nvidia. In the face of import restrictions, the CCP is exploring an additional $30-70B commitment from the government (in addition to their technology investment funds), outdoing the US government’s allocation from the 2022 CHIPs act. Chinese chip designers like Moore Threads, MetaX, Cambricon and Huawei will race to achieve this national objective.

In addition, jailbroken AI coding models will also power multiple critical cyber attacks in 2026. As the scale and sophistication of coding models improve, the sophistication and manpower required to launch a powerful cyberattack will drop.

Bad actors will also be able to operate at much greater scale and with greater efficacy than ever before. Anthropic's November 2025 report regarding a Chinese state-sponsored hacking group jailbreaking Claude and launching a September attack on 30+ US corporations is a major evolution from the "vibe hacking" that Anthropic traced back in August.

Claude agents were used for target selection, reconnaissance and surface mapping, vulnerability discovery and information extraction. Without meaningful investment in cyber defenses in 2026, AI will tip the scale in favor of attackers for this calendar year.

AI Agents Will Redefine “Human-Only” Workflows

Contributed by Mendy Yang ([email protected])

AI is reshaping processes that we’ve historically assumed had to be “human only.” We’re already seeing early examples like AI recruiters running full interview cycles that show how much better these experiences can become when we remove scheduling friction, ask more nuanced questions, and eliminate ghosting.

Now is the time to rethink these overlooked, high-touch workflows and rebuild them with AI from day one. AI agents will continue outperforming humans in end-to-end workflows that people once assumed needed a person in the loop.

This won’t happen just because AI is cheaper, but because the experience will be faster, more consistent, and fundamentally better for users. I’m looking for teams that spot these opportunities early and build tools that catalyze behavioral change before it becomes standard.

The Next Generation of Breakout Companies Won’t Come From “Central Casting”

Contributed by Aaron Holiday ([email protected])

Gen-AI coding tools like Replit, Loveable, and Cursor are collapsing the distance between inspiration and execution, allowing founders to ship MVPs at near-instant speed. Combined with the labor force disruption caused by AI adoption across industries, this will empower a new generation of entrepreneurs ranging from high school students to employee spinouts who previously lacked the engineering and product support or confidence to take the initial leap.

At 645, we’re especially energized by the rise of these non-traditional, non-“central casting” founders who will launch the next wave of breakout tech companies. In 2026, we expect to see a surge in startup formation and new product releases at a scale the industry has never witnessed, driven by the unprecedented fluidity from idea to software product, now possible even without technical knowledge.

About 645 Ventures: We back founders bringing the invisible to life. With $550M+ under management, we partner with top institutional investors to support companies poised for industry transformation and large-scale M&A or IPO events. We have made over 120 investments in our history, including companies such as Iterable, Shift5, True Anomaly, Setpoint, Overtime, Squire, and ShopCircle. We invest at the Seed, Series A, and growth stages. Learn more at www.645ventures.com.