With every new frontier model—GPT-4o, Claude 3.5, Gemini Ultra—the cost of intelligence is falling fast, even as capability keeps rising. We’ve seen this dynamic before: when a broadly useful resource becomes dramatically cheaper and more powerful, it triggers a wave of experimentation. Cheaper transistors gave us the PC era; cheaper bandwidth gave us the web. Now, as intelligence becomes a low-cost input, we’re seeing it embedded everywhere—in tools, workflows, and physical systems. The result is a flood of new products, new companies, and eventually, a reshaping of the broader economy. We’re at the start of that shift.
So when intelligence gets cheap, what changes? Some things become newly scarce (GPUs, energy, frontier models). Others become newly essential (AI infrastructure, data, safety). Entire domains become newly executable by AI (research, creation, knowledge work, physical tasks).
Along these lines, below is the landscape we believe holds the next generation of outsized returns.
Lower-Cost, Powerful Intelligence: First-Order Effects
Energy and Compute
Running advanced AI models at scale requires enormous electricity and high-end semiconductors; thus, companies controlling energy-efficient compute infrastructure, advanced chip IP, and raw energy production could see sustained demand. In this emerging economy, primary assets may shift from traditional capital to flows of energy and computation.
Frontier Model Companies
These are deep research labs building foundation models at the frontier. It takes world-class talent, massive amounts of compute, and billions in capital. The field is consolidating quickly. But when it works, and the model becomes a platform (e.g., ChatGPT), the upside is enormous.
AI Infrastructure
Old model: Build tools and infra for developers to create next-gen apps like AWS, Snowflake, Confluent, and GitHub.
New model: Build infrastructure that enables AI agents to operate autonomously through software that navigates the web, pulls real-time data, takes action, and completes tasks.
To function in the real world, AI agents need a new stack of foundational layers:
- Execution environments — where agents run, reason, and act
- Data layers — real-time, structured context for grounding decisions
- Observability & safety — to monitor behavior and ensure reliability
This is the new runtime for work. Infrastructure built not for human users, but for AI agents operating independently.
AI for Scientific Discovery
AlphaFold showed what’s possible when AI is applied to hard science, unlocking biology problems that had stalled for decades. AlphaDev, AlphaTensor, and AlphaChip are doing the same for algorithms and hardware. This is the next wave: AI as a tool for invention, not just automation. These companies are vertically integrated with domain-expert teams and blending deep science with cutting-edge AI. We believe this is where some of the most important long-term AI breakthroughs will come from.
Co-Pilot for X
We’re seeing vertical AI tools emerge across workflows, including coding, design, CAD, email, finance, legal, and more. Most start as co-pilots, but over time, they become more agentic, going from suggestion to execution. As this shift happens, the user’s role evolves from operator to supervisor. The software becomes the worker, and the human becomes the manager.
AI as Labor
AI agents are replacing humans in repetitive, rules-based tasks across a range of functions. AI SDRs are managing outbound sales prospecting, messaging, and follow-up. AI triages alerts and escalates threats automatically for security operations. AI agents flag fraud, investigate transactions, and generate case files for money laundering. AI simulates attacker behavior to detect abuse and stress-test defenses at scale for red teaming. These aren’t co-pilots, they’re full human replacements. The shift is from human execution to AI execution with the user overseeing the output. Much of the BPO industry will be substituted by AI. Call centers, claims processing, onboarding, and data entry are large-scale, repetitive workflows that AI is well-suited to handle. As agent infrastructure matures, the substitution curve will steepen.
Robotics
As AGI-level reasoning meets falling hardware costs, physical robots will rapidly expand from controlled factory floors into dynamic human environments: warehouses, retail stores, hospitals, restaurants, farms, and even homes. The combination of dexterous manipulators, vision-language models, and fleet-level learning enables robots to handle unstructured tasks—picking mixed-SKU inventory, stocking shelves, preparing food, harvesting crops, delivering packages, and assisting the elderly. The economic effect mirrors cloud computing for physical labor: cap-ex becomes op-ex as companies subscribe to robot fleets that improve continuously via shared data. The winners will bundle hardware, autonomy software, and vertical workflow integrations, capturing recurring “labor-as-a-service” revenue while compounding data moats.
Second-Order Effects
But those are just the first-order effects—where AI is being built and deployed. As the cost of intelligence keeps falling and more work shifts to machines, second-order effects will start to take hold. Time will free up. Work routines will compress. And human attention will flow toward what still feels scarce: experiences, connection, well-being, and creative expression.
We’ve seen hints of what happens when work recedes. During Britain’s energy crisis in the 1970s, a three-day workweek led to a surprising boom in leisure—golf courses filled up, fishing shops sold out, and late-night radio audiences tripled.
These are the non-software markets we think will matter more—not despite AI, but because of it.
Tourism & Travel
Demand could surge for authentic adventures (like remote mountain treks or immersive cultural tours), while AI handles routine mass tourism (think robot-piloted planes or automated check-ins). Popular destinations may raise fees to preserve natural beauty, prompting tourism boards to emphasize genuine human elements like local guides and artisan souvenirs.
Entertainment
The appetite for live events, interactive experiences, and human-led creations will grow in a world where AI-generated content is abundant and can simulate everything. While some entertainment ventures embrace AI to produce on-demand shows or immersive games, businesses that highlight unique storytelling, genuine artistic talent, and communal experiences are poised to stand out.
Wellness & Mental Health
With AI doctors capable of real-time triage, personalized fitness coaching, and telemedicine, the wellness industry expands to emphasize preventive care, holistic health, and AI-assisted therapies, letting people embrace group fitness, nature retreats, and lifelong self-improvement. Over time, mental health support weaves into everyday life, fostering emotional resilience and purposeful growth on a broad scale.
Food, Drink & Hospitality
Automation in agriculture and cooking makes basic meals cheap and abundant, freeing people from routine food prep. However, dining out and hotels may split between ultra-automated efficiency and premium experiences with human touches, appealing to those craving genuine social rituals.
Arts & Collectibles
AGI may churn out digital art, but pieces created by human hands, steeped in cultural heritage, could become new “blue chip” assets for collectors and investors. As production becomes more automated, people will pay a premium for limited-edition or physically scarce items, driving niche markets and collector communities.
Education
AI tutors and adaptive platforms handle routine instruction, allowing human teachers to focus on mentoring, creativity, and emotional support. Schools emphasize critical thinking and collaboration, making lifelong learning the norm in a swiftly changing world. And a humanities renaissance is certainly a possibility.
Local Advocacy
Communities could flourish as automation frees up time for grassroots activism, with more people engaging in civic projects, volunteering, and local decision-making. This heightened participation strengthens social bonds, empowering citizens to shape their neighborhoods in tangible ways.
