2026: The "solar terms" in the AI industry have changed; how can entrepreneurs fine-tune their "algorithms"?

As the new year begins, I want to share some thoughts for 2026.

At the turn of the year, events like Meta’s acquisition of Manus and the IPOs of Zhipu and Minimax on the Hong Kong Stock Exchange have delivered a powerful boost to global AI entrepreneurs, validating the immense opportunities of our era with real capital.

From my perspective, the long, anxious “waiting for spring” phase in AI entrepreneurship over the past three years has come to a close.

Why do we still need to “wait” for spring? It’s similar to China’s traditional 24 solar terms—each is suited for planting different crops. If ChatGPT marked the “Beginning of Spring,” then after passing 2024’s “Rain Water” and seeing 2025’s “Awakening of Insects,” 2026 could well be the year of the “Spring Equinox.”

Looking back at the pace of entrepreneurs and the capital markets, this view is gaining support. In 2023, most domestic AI entrepreneurs who secured funding were building large models. In 2024, many began experimenting with so-called “shell” applications, exploring new directions, but capital remained cautious and few entrepreneurs achieved broad consensus. By 2025, applications solving real problems started to emerge, and domestic capital became more active.

For example, top domestic VCs were generally cautious in 2024, but some institutions made dozens of investments in 2025, signaling that the accelerator is now pressed down about 80%. This “hard acceleration” is becoming mainstream, and the speed of investment consensus has noticeably increased—mainstream firms are now focusing on AI hardware and moving quickly.

By 2026, large model capabilities will continue to accelerate. From a capital perspective, recent events like Manus’s acquisition and the IPOs of Zhipu and Minimax are landmark triggers for chain reactions in decision-making, and enthusiasm for “full throttle” investment will rise sharply. Changes in technology, capital, market, and user landscapes are all creating a shift in the “seasons.”

So, let’s discuss a direct question: In 2026, what kind of entrepreneurs will most easily secure funding and make significant progress?

From “Interesting Exploration” to “Meaningful Delivery”

Two years ago, a cool demo might ignite user interest and even attract investment. Today, a “cool” product is no longer enough—it must be genuinely useful, even overwhelmingly superior to previous generations, or introduce a new definition for its product category.

This shift is evident not only in AI software applications but also in hardware products showcased at this year’s CES.

In previous years at CES, an AI-branded product with a few attractive features could spark huge attention and excitement. This year, you can clearly sense the shift—people are no longer paying for mere “AI concepts” or superficial “AI features.” Continuing to treat AI as a “cosmetic”—simply enabling AI conversations or basic AIGC features—is unlikely to succeed.

The industry and users are returning to rationality: AI should not be just a superficial “cosmetic”; it must be the “skeleton” supporting the product.

Today, what truly excites people is not whether a device is “AI hardware,” but whether it is the best in its chosen scenario. This depends on whether, in a sufficiently defined context, AI—regardless of whether large models are used—is integrated as an engine or foundational capability to deliver a clearer, higher-quality user experience and value than ever before.

These deeper questions about value are becoming the focus for investors and increasingly sophisticated users.

To achieve this shift from “cosmetic embellishment” to “structural support,” we must re-examine the entry points for product and AI capabilities. In this “season,” choosing “specialized” over “generalized” approaches will clearly lead to more meaningful delivery.

Generalized products offer users unlimited possibilities and capabilities, expecting them to explore “what can be done.” However, without clearly defined scenarios, most users will feel lost, and the product may miss its chance to connect during initial interactions.

Specialized products allow you to focus on solving specific problems for targeted groups, aligning with their needs during development and marketing, and concentrating resources for more meaningful delivery. With the rising capabilities of AI and your deep understanding of the scenario, combined with effective organization of AI abilities, you are more likely to wow users at first sight.

This specialized approach helps you build strong user engagement and data feedback loops from the start, making it easier to establish a foothold compared to general platforms, with potential for future horizontal expansion.

At this stage, entrepreneurs need to ask: Does the problem you aim to solve truly exist? And does your solution offer an overwhelming advantage? These two questions become especially important.

The Exciting “Extension Line”

Of course, the more focused the scenario, the clearer it is, but the ceiling for value may not seem very high. If you only aim to be an independent developer, taking a solid first step is enough. However, if you want VC funding to fuel a scalable, high-ceiling company, you must consider a second question: Where is the future extension line? Or, what is the ultimate goal?

In conversations with many investors and in my own project reviews, an entrepreneur’s ability to capture a small entry point represents the “baseline,” while the potential for future extension from that entry point represents the “ceiling.” A credible extension line is crucial at this stage.

This extension line isn’t about telling a story; it’s often embedded in your initial design. We can break it down into two dimensions:

1. Don’t just look at ARR and sales; consider what data assets can grow inside your product “container.”

In the AI era, if your product’s core capability can’t strengthen with user engagement, how is it different from traditional software? ARR, a concept from the traditional software era, is a good SaaS PMF validation, but it doesn’t fully reflect the long-term value of AI-era products.

Many entrepreneurs I’ve met recognize that AI-era products are essentially self-propelling “growth containers.” There’s little debate about whether to integrate model and product—a strong product company will eventually have its own model and become a “model company.” But before that, one core task is to become a “data company,” building product moats with fresh, live data sourced from user needs. This self-reinforcing feedback loop is key to moving from narrow to broad scenarios, from low to high LTV (lifetime value), and driving sustained commercial growth.

1. Your “supply chain” shouldn’t be “flat.”

Whether software or hardware, supply chain issues are worth examining today. Simply calling large model capabilities or leveraging China’s supply chain advantages to build demos is possible, but true products require a supply chain that’s streamlined but sophisticated. From the outset, consider how to build a meaningful supply chain—such as unique engineering assets (domain-specific labeled data and models, workflows, and data accumulation), or, when resources allow, immediately customizing and strengthening general supply chains (as DJI and Unitree Robotics did with motors). The deeper your supply chain, the less likely you are to be easily copied, and the longer and greater your lead.

In a very “flat” supply chain, you control few links, and your differentiation may be limited to UI, ID (industrial design), and a bit of first-mover advantage. In a supply chain you personally extend, you can create “high ground.” For example, leveraging user data insights for personalization, abstracting workflows to lower inference costs, or adapting computing, sensing, optics, and local models for better fit—your product is the visible tip of the iceberg, while your supply chain is the massive part below the surface. The larger the submerged portion, the bigger the visible part above water.

From today into the future, from the first step to the ultimate goal, clearer thinking and choices mean greater local intensity, and also a freer strategic path.

Entrepreneurs in 2023 may have said, “Just start doing!” but by 2026, asking “How should we do it?” becomes the critical prerequisite.

Geekpark published an in-depth review of DJI’s technical DNA (In-depth Review: How DJI Became a New Giant in Imaging). Years ago, Wang Tao defined the core value of drones as “a flying camera,” which became both the starting point and future direction, leading to deep capability stacks beyond flight control, including motors, gimbals, and imaging. This imaging-focused “extension line” ultimately shaped DJI into not only the undisputed leader in drones, but also a major player in the new era of imaging, with products like the Pocket 3 achieving sales far beyond consumer drones.

Great Entrepreneurship Is Often a “Three-Step Layup”

If we compare entrepreneurship to basketball, the “hoop” is the ultimate user value you aim to create.

You can choose the thrill of a “half-court three-pointer,” or the agility of a “three-step layup”—both are paths toward the same goal.

The “half-court three-pointer” means aiming for a grand direction and shooting for the final goal from the start, hoping it lands. Entrepreneurs of this type often have the credentials to attract $100–200 million in funding right away, with a “basket of balls” at their side. If one shot misses, capital and resources allow for several more attempts. Of course, don’t envy them too much—the bigger the expectation, the greater the pressure; each path has its own challenges.

For most “ordinary entrepreneurs,” you may only have “one ball.” Without hundreds of millions in margin for error, your most rational strategy may be the “three-step layup.”

The “three-step layup” requires you to find a solid demand and scenario entry point under the given technical and resource conditions, build a streamlined yet sophisticated supply chain, and ensure every product appears to have no barriers yet actually has high barriers. The ceiling may seem limited, but you know it’s a continuous breakthrough connecting the dots into a line.

Recently, Manus was acquired, and many thought this was a “half-court three-pointer”—but that’s a misunderstanding. Over a longer time frame, Manus’s success is actually a classic “three-step layup.”

Manus founder Xiao Hong didn’t start with huge capital to build a general agent. His first step was Monica, a lightweight browser plugin that quickly validated PMF at low cost and captured real user needs in the AI era. When he realized the browser plugin path had limited room for advancement, he decisively took the second step—leveraging previous accumulation to pivot to the broader general agent field with Manus, investing heavily in engineering to extend and deepen the supply chain, bridging countless gaps between models and complex application scenarios, and pioneering agent user value delivery.

Then, benefiting from “first-mover advantage,” user behaviors (such as concentrated task needs) helped Manus focus the general architecture on more vertical tasks with “effective delivery.” This focus brought more users and revenue, and at peak momentum, the acquisition was completed—scoring the “basket.” This is a textbook “three-step layup.”

Is it possible for an entrepreneur with a “natural” $100 million funding to avoid half-court shots and opt for the three-step layup, reducing fundraising costs and exploration pressure, while quickly iterating and shortening the distance to the hoop to improve innovation success rates?

Absolutely! These may be the most formidable players.

There is no universally perfect strategy for entrepreneurs—only the best fit for your resources, technical stage, and market environment. You can learn from history, but history never repeats itself exactly. 2026 may not produce another Manus-style story, but it will certainly deliver more exciting new ones.

I believe that the just-ended 2025 will be the toughest year for AI entrepreneurship in the next five years. In different seasons, sow different seeds, write different stories.

Wishing everyone finds their own “hoop” and takes action in the new year.

Statement:

1. This article is reprinted from [Geekpark]. Copyright belongs to the original author [Zhang Peng]. If you have concerns about the reprint, please contact the Gate Learn team for prompt processing in accordance with relevant procedures.
2. Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute investment advice.
3. Other language versions of this article are translated by the Gate Learn team. Unless Gate is mentioned, do not copy, distribute, or plagiarize translated articles.