a16z: AI New Cycle, Robots Rise, Hardware Surges, Software Reverses

a16z notes that AI is driving capital from virtual to physical, robot investments hit new highs, and enterprises need to break through the exploration bottleneck to adopt AI.
(Previous context: Anthropic former researcher founded Mirendil, raising $200 million, claiming to be an "AI that can self-upgrade")
(Background supplement: Employees burned $80k in tokens to create a "meme shooter game" using AI, and the boss went viral asking everyone to play after seeing the bill)

Table of Contents

Toggle

• Cycle Reversal: Tech Winners Become Losers
• Physical Buying Spree: Robots Surge 4.5x
• AI Transformation: Accenture Plunges 6x
• Deep Restructuring: AI Use Cases Surge 44%
• Lean Operations: AI Venture Investments Hit New Highs
• Solo Entrepreneurship: Million-Dollar Revenue Doubles

Recently, a16z analyzed the core trends of the current tech and business cycles from multiple dimensions including market investment, AI applications, entrepreneurship ecosystem, and retail industry. The article points out that driven by the AI wave, capital markets are gradually shifting from a preference for light-asset, consumer internet to hardware, robotics, and other physical industries. At the same time, AI is reshaping corporate organization, entrepreneurial barriers, and productivity growth logic. Details follow.

If you compare this cycle with the previous one, you'll notice that in some ways they are identical, while in others they are completely opposite.

The similarity is that in both the post-financial crisis period (2010-2020) and the post-pandemic period (2020-present), the tech industry has been the winner of the cycle. However, the landscape of other industries has changed dramatically: the winners of the previous cycle have become losers, and vice versa.

• Healthcare, consumer goods, and media industries all achieved double-digit returns after the financial crisis, but now they hover around 3% to 6%.
• Meanwhile, returns in energy, raw materials, construction, and finance have risen from low single digits to mid-to-high double digits.

Industries that once lagged have become leaders, and the leaders have become laggards.

The tech industry is an exception, having always been a cycle winner, but there are nuances. Hardware is the real standout this cycle (it performed quite well last cycle too), but software follows the overall reversal trend.

Stepping back, there is a very obvious pattern mentioned before: the market has shifted its attention from light-asset, consumer-oriented industries to the heavy-asset "physical" economy, largely driven by AI infrastructure buildout.

This is a rotation from bits (virtual) to atoms (physical).

"Heavy-asset" companies have turned the tide after lagging "light-asset" companies for over a decade.

Of course, if this cycle resembles the previous one, the overall trend is that all this heavy-asset infrastructure will ultimately expand into the software/application layer. In the post-financial crisis era, chip manufacturers (and cloud service providers) dominated early, but eventually gave way to applications, markets, and enterprise software that flourished on cloud platforms (driven by chips) powering phones, computers, and servers. In other words, the shift to the virtual layer was temporary and cyclical, not a more lasting structural shift.

Cycle Reversal: Tech Winners Become Losers

This could certainly happen again this time: in fact, if the AI infrastructure buildout does not eventually expand into the light-asset layer, it could be quite disappointing (both could also develop in tandem). Even so, in the public markets, some signs reveal that the "virtual revolution" may have its own sustainability. And strictly speaking, it's not just about AI infrastructure.

The premium for "real-world" technology is emerging in private markets, not only in AI infrastructure but also in robotics:

Measured by market cap of the top 100 private companies (by category), robotics (and physical AI) didn't even appear on the list in 2016, but a decade later, it has surpassed fintech and payments to become the second-largest category.

If you look at venture capital flows, you'll also see a surge in interest in robotics:

According to PitchBook data, investment in robotics and physical AI hit all-time highs in Q1 in both dollar amount and deal count, with approximately $16 billion invested across nearly 500 deals.

For reference, the robotics investment boom is about 2x higher in deal count and about 4.5x higher in value compared to the 2021-2025 period.

The key point is that the shift to the physical economy (at least in private markets) is not just about chips and inference: hardware as an independent product is rising.

This is not hard to grasp either. Better software has enormous potential, but robotics pushes technology into a range of real-world "tasks" that software alone cannot reach. AI in some ways unlock software that drives hardware, expanding the scope of demand in unprecedented ways. This is quite similar to how electricity eventually enabled machines to perform work that humans could hardly imagine.

Currently, the most striking new area for robotics is defense. Of course, the growing global defense budget also plays a role. If all goes according to plan, the shift toward asset-intensive industries may be deeper, broader, and more enduring than any modern tech cycle before.

Physical Buying Spree: Robots Surge 4.5x

In the early stages of the large language model (LLM) wave, management consulting firms were identified as potential winners in the AI space, at least in the short term. The logic was simple: enterprises want to use AI, so they will hire consultants to study how to do it. Accenture in particular was noted to be in a very favorable position because it could not only provide advice and roadmaps but also offer end-to-end services, so-called "managed services."

For reasons that may vary, market optimism about Accenture seems to have evaporated:

Accenture's free cash flow multiple spiked as high as 30x in early 2025 but has since fallen to about 6x, roughly one-third of its long-term average.

As for why the market lost confidence in Accenture so quickly, you can judge for yourself. But one thing is becoming increasingly clear: in the broader field of "adopting AI," the meaning goes far beyond simply adopting AI. Not all AI adoption creates value in the same way, and truly (or more effectively) adopting AI, according to some recent research, requires subtle strategies in the development and ideation phases.

In a study involving 515 high-growth startups, researchers focused on what it really means to be a "native AI" company. More specifically, they wanted to know how to transition from "AI improving tasks" to "AI improving the company," and the results were quite striking.

It turns out the key lies in what researchers call the "mapping" problem.

When the companies in the study were told how other companies were reorganizing production around AI (the "experimental group"), they embarked on a distinctly different exploration process. Instead of simply replicating existing processes, the experimental group started further upstream, integrating AI into business outcomes, thus forming entirely different processes.

Researchers used product development as an example:

In this case, AI did not replicate existing steps in the process but redesigned the process around its own capabilities, even though the goal was to achieve the same basic business outcome.

AI Transformation: Accenture Plunges 6x

Of course, this is just one example, but overall, the productivity impact of AI on "experimental group companies" is enormous. Experimental group companies:

• AI use cases increased by about 44%:
• Revenue for the top 5% of companies was about 2x (top 10% saw a 50% increase):
• Capital consumption decreased by about 40% (the gap was larger at both ends of the distribution).

In short, when high-growth startups truly set out to "adopt AI," they observed more application scenarios, generated more revenue, and consumed less capital than those that didn't.

This is a rather astonishing result, which can both alleviate some concerns about the "AI ROI problem" and explain why AI's ROI has not yet fully manifested at the enterprise level, at least not to the extent some expected.

Researchers point out that this means: (a) AI-driven productivity gains at the enterprise level are indeed transformative; but (b) the real breakthrough lies in the exploration phase, meaning that "exploring where and how to deploy AI is the key bottleneck to realizing gains," and this is not simply about "adopting AI."

In this sense, the existence of an "exploration bottleneck" means that AI's development path is no different from previous technology-driven productivity leaps.

For example, when electrification first became widespread, many manufacturers simply replaced steam engines with large electric motors while retaining the original overhead shafts and belt drive systems. The factory was basically unchanged, just "with electric motors this time." However, real benefits only began to emerge when manufacturers realized they could install small motors on each machine (and almost completely abandon the entire shaft and belt drive system): factories were eventually redesigned around embedded electric power (rather than the reverse). And of course, what followed was one of the great milestones in the history of productivity leaps.

Regarding AI, startups, and academic research, the same group of researchers also observed one more thing: AI startups are indeed streamlining operations. At least according to this study based on YC startup data from the past four years.

Deep Restructuring: AI Use Cases Surge 44%

Researchers examined data from YC W20-F24 cohorts (first funding rounds completed between 2020 and 2024) and correlated it with Revelio's employee count, function, and seniority data. They wanted to understand whether AI startups differed from non-AI startups in hiring and/or organizational structure.

They observed:

• AI startups start smaller and operate smaller:
• The distribution of startups with fewer employees is heavily skewed toward AI startups:
• AI startups tend to have flatter hierarchies, with the largest proportion of AI startups in companies with few or no layers:

The implications are clear, although there may be more variables in the details. But you get the idea: if you argue that AI will enable companies to create more value with fewer resources, this survey certainly provides more evidence for your view.

Additionally, Stripe Economics has again weighed in on the "solo entrepreneur" phase enabled by AI.

(Note: Recently, Ernie Tedeschi of Stripe Economics, based on Stripe's own data, pointed out that all types of founders seemed to grow in Q1, but "non-AI solo founders" grew the most, with "AI solo founders" coming in second. As shown below:)

Although Stripe points out many limitations in how they identify "solo entrepreneurs" in their data, they still provide more evidence for the view that AI is indeed driving more entrepreneurial activity and business formation, and that solo entrepreneurs are achieving considerable success.

Take a look at the proportion of solo entrepreneurs by revenue threshold:

Lean Operations: AI Venture Investments Hit New Highs

Not only is the proportion of solo entrepreneurs with annual revenue over $100k rising, but the proportion with revenue over $5 million and $10 million began to grow significantly in 2023 and 2024.

Stripe Economics notes:

• We observe a significant increase in the number of solo entrepreneurs with annual revenue over $100k in our index, but the number with revenue reaching higher thresholds has grown even more, with a noticeably accelerated pace since 2023. In 2025, the number of solo entrepreneurs with revenue over $1 million is more than double that of 2023, while those with revenue over $5 million and $10 million are nearly triple the 2023 figures.
• Perhaps more interestingly, the proportion of solo entrepreneurs exceeding these revenue thresholds has also doubled over the past two years. This suggests that the surge in business formation does not reflect low-quality experiments by a few lucky individuals, but rather that the new cohort of solo entrepreneurs may be of higher quality than ever before.

Of course, given the uncertainties around how solo entrepreneurs are identified (in this case, through Stripe's solo entrepreneur-specific tools) and the possibility that these businesses' employee counts may change over time (which Stripe may not be aware of), the data indicates that the AI-driven era of small businesses is still evolving.

One interesting aspect of grocery stores is that, unlike the broader retail trade category, grocery stores have not seen significant productivity gains over the past 30 years:

Or more precisely, since 1990, retail productivity growth has remained fairly steady, while grocery store productivity first declined, then recovered somewhat, then stagnated, and though it has recently declined again, it has started to recover, but it remains far below the retail productivity surge.

This is interesting because it tells both the story of technology (and its relationship to productivity) and the story of how productivity is measured, roughly as output divided by labor hours (which is at best an imperfect measure).

For grocery stores (and retail), aside from the cash register, the greatest invention was the electronic scanner. They first appeared in the 1970s, but by the 1990s, they were nearly ubiquitous. Scanners served two main purposes: (1) greatly expanding inventory variety; and (2) enabling increasingly granular data collection by retailers and grocers to understand customer willingness to buy and the necessary inventory levels.

In the 1990s, both grocery stores and retailers began to expand significantly in scale, benefiting from technology-driven economies of scale, which was good for consumers but more or less signaled the end of the family-run mom-and-pop store.

From that point onward, however, the fortunes of retailers and grocers diverged. Retailers greatly expanded inventory but did not add many new employees, instead focusing more on ready-made pre-packaged goods that could be managed and monitored with far fewer people than before. Grocers, on the other hand, decided to expand into specialty services beyond groceries, such as florists, bakeries, deli counters, etc.

Solo Entrepreneurship: Million-Dollar Revenue Doubles

Of course, as the share of specialty services grew, the demand for specialized labor increased. As shown in the figure above, although grocery store productivity improved in terms of greatly expanding the variety of goods and services and lowering prices, their "productivity" as measured by output per labor hour did not increase. This is also why retail "productivity" far exceeds grocery "productivity," while wage growth for both has been roughly the same.

​​It wasn't until grocers adopted successful practices from the broader retail and department store industry that their productivity began to rise again:

Around 2000, the share of non-household food products began to grow significantly: higher-margin pre-packaged foods, snacks, and general merchandise grew nearly 5-fold within a decade. At the same time, supermarkets outsourced more stocking and display tasks to suppliers, similar to charging "slotting fees" for shelf space. This was a clever "productivity-enhancing" strategy: while labor hours did not decrease, they were shifted to others.

From a "productivity improvement" perspective, this shift increased output without increasing labor hours, leading to a revival in supermarket productivity.

Although labor's share of grocery revenue rose steadily until around 2002 (while labor's share of retail revenue declined), both shares have been steadily declining at least until recently.

The decline in "labor share of income" is essentially the flip side of "productivity": producing more with fewer workers leads to a decline in labor's share of income (not accounting for the growth of 401k retirement accounts from all the profits).

Interestingly, however, (returning to the topic of technology and productivity), the latest wave of shopping innovation (e-commerce and delivery) seems to once again coincide with the divergence in "productivity" trends between groceries and retail. While e-commerce has been a boon for retailers, who no longer need to lease physical stores, delivery may simply mean that the same or more people shop around and pick items in grocery stores. Curbside pickup can even be more labor-intensive than traditional shopping.

Whether this is causation or coincidence, the fact is that after the pandemic, grocery productivity has declined again (and labor's share has started to rise), while retail has become leaner and more efficient. The same technology, the same productivity gains, yet the resulting "productivity" is vastly different.

However, the good news for grocers is that advertising on the floor always brings in money (high margins).