The Remarkable AGI Trades of Daniel Gross
Original author: @johncoogan
Peggy, BlockBeats
Editor’s Note: At the beginning of 2024, AI was still in a phase marked by both hype and uncertainty. At that time, Daniel Gross posed 18 questions on a single page: Where will value flow? Will energy become a bottleneck? Will software engineers be replaced? How will the competitive landscape between nations change?
Looking back two years later, these very questions are more insightful than any specific prediction. AI profits have indeed been concentrated at the infrastructure layer—NVIDIA emerged as the biggest winner; energy and power have rapidly become new strategic bottlenecks; API costs have plummeted, while compute, capital, and geopolitical risks continue to escalate.
This article revisits the key questions Gross raised at the time and examines them one by one against the real-world developments of the past two years. It is not only a retrospective on the investment logic of AI, but also a roadmap for understanding how technological revolutions are reshaping market structures, industrial chains, and the global power landscape.
The following is the original text:
In January 2024, Daniel Gross, then CEO of Safe Superintelligence and now head of product at Meta AI, published an article titled “AGI Trades.”
This article, spanning just one page, lists a series of questions about the potential impacts of AI advancements. Looking back more than two years later, these questions appear remarkably prescient, even though no definitive conclusions were provided for each at the time. Below, we revisit each of the 18 questions he raised.
Markets
In a post-AGI world, where will value flow?
Currently, value is indeed concentrated at the infrastructure layer—chips, packaging, power, and related areas. NVIDIA has captured over 100% of the profits from the AI boom, as many other companies are still operating at a loss. This is clearly reflected in market capitalization changes: NVIDIA’s market cap has increased by $3.2 trillion, rising from $1.2 trillion to $4.4 trillion; in comparison, cloud platforms have seen much more modest gains (Microsoft up 4%, Amazon up 30%).
On the private market, the valuations of OpenAI, Anthropic, and xAI have also grown dramatically; however, their combined $1.4 trillion increase in value still falls short of NVIDIA’s market capitalization growth during the same period.
This is a crucial issue at the very beginning of 2024.
What will happen with NVIDIA and Microsoft?
NVIDIA has performed exceptionally strongly, with revenue increasing from $60.9 billion in fiscal year 2024 to $215.9 billion in fiscal year 2026, nearly tripling.
Microsoft, however, does not hold as strong an advantage. Although Azure’s growth accelerated to a 40% year-over-year rate, Microsoft’s stock price rose only 4% from January 2024 to March 2026. The market has questioned its annual AI capital expenditures exceeding $80 billion—how and when these investments will translate into returns remains unclear.
In this AI gold rush where companies are selling shovels and picks, NVIDIA is clearly the biggest winner, while Microsoft’s bets on infrastructure have yet to deliver noticeable returns to shareholders.
Is copper mispriced?
Indeed, it has been severely undervalued. In January 2024, copper was priced at $3.75 per pound, rising to a record high of $6.61 per pound two years later.
AI has an enormous demand for copper. For example, the NVIDIA GB200 NVL72 server rack uses over 5,000 copper wires; if laid end to end, they would stretch more than 2 miles. A 100MW data center requires approximately 3,000 tons of copper.
Overall, data centers may consume 500,000 tons of copper annually. Some therefore say, “Copper is the new oil.” Of course, many other things have also been called “the new oil,” as AI infrastructure is extremely complex, with bottlenecks existing in nearly every stage. Thus, this claim should be viewed with caution.
Real Estate
If AI can write all software, will San Francisco become the new Detroit?
It depends on what is meant by "the new Detroit."
AI actually saved San Francisco, preventing it from becoming a declining city like Detroit. Today, San Francisco continues to thrive:
Office vacancy rates decreased from 36.9% to 33.5%.
OpenAI owns 1 million square feet of office space.
Anthropic owns a 25-story office building.
Sierra has signed a 300,000-square-foot office space.
In the first half of 2025, 78% of U.S. AI venture capital funding flowed to the Bay Area. Of course, there is another side: overall employment in San Francisco remains below pre-pandemic levels, yet housing prices have stayed strong. Therefore, it is certainly not a "ghost city." The urban environment has also become cleaner.
How will AI affect wealth inequality?
It's still too early to draw conclusions, as the data changes are not yet significant, but some studies are already worth noting.
The IMF’s 2025 study suggests that AI may reduce wage inequality (by automating high-income jobs), but could worsen wealth inequality (as capital gains concentrate among tech company owners). OECD research found that wages for low-skill jobs grew the fastest (assembly workers +11.6%), while wages for high-skill jobs grew the slowest (CEOs +2.7%); however, this may reflect minimum wage policies more than AI itself.
Concentration is also rising in capital markets: the "Magnificent Seven" account for approximately 32% of the S&P 500’s market capitalization and contributed about 42% of total returns in 2025; meanwhile, massive funding rounds for AI startups—such as OpenAI’s $110 billion and Anthropic’s $30 billion—have generated enormous private wealth for a select few founders and investors.
Energy and Data Centers
How should you invest if AI becomes an energy competition?
This judgment is entirely correct. AI has indeed become an energy game.
Those who caught this trade made a lot of money. For example:
- Vistra: +321%, the second-largest gain on the S&P in 2024 (after Palantir)
- Constellation Energy: Stock price has tripled since the release of ChatGPT
- NRG Energy: Approximately 95% single-year increase in 2025
- Oklo: Over 700% increase in 12 months
Nuclear energy is experiencing a surge:
- Microsoft signs a $16 billion, 20-year PPA to restart the Three Mile Island nuclear plant.
- Google signs a 500 MW small modular nuclear reactor (SMR) agreement with Kairos Power
- Meta signs power purchase agreements for 6.6 GW of nuclear energy with multiple nuclear companies.
Energy has become one of the most successful investment themes of the AI era.
Which parts of the data center supply chain are hardest to scale by 10x?
The bottleneck in the chip industry is CoWoS packaging (TSMC’s Chip-on-Wafer-on-Substrate).
In the data center industry, the biggest bottleneck may be the power transformer.
- Delivery cycle is close to 3 years
- A 30% supply deficit is expected in 2025
Costs have increased by 150% since 2020.
This 100-year-old technology is the key bottleneck limiting how quickly data centers can connect to the power grid.
Is coal undervalued?
To some extent, yes, but far less than copper. Coal prices actually fell by about 22% in 2025 and rebounded slightly by early 2026.
Coal companies performed reasonably well:
- Peabody Energy: +34%
- CONSOL Energy: +37%
Meanwhile, U.S. coal-fired power generation increased by 13% by September 2025.
States with rapid data center growth are particularly notable:
- Ohio: +23%
- Oklahoma: +58%
Nations
Who are the winners and who are the losers?
The clear winner is the United States.
In 2024, private AI investment in the United States reached $109 billion (compared to just $9.3 billion in China), with cumulative investment since 2013 totaling $470 billion—exceeding the combined total of all other countries. In 2024, the United States launched 40 significant AI models, while China launched 15.
The game isn't over yet, but for now, the United States is the center of the AI competition.
What would happen if India's $250 billion GDP export depended on GPT-4 tokens?
The situation has begun to emerge but is still in its early stages. Hiring in India’s IT outsourcing industry has noticeably declined. Between 2024 and 2025, major IT companies laid off approximately 58,000 employees, whereas the industry added 360,000 workers between 2021 and 2023.
Will software engineers be replaced like typists in history?
Software engineers have not yet taken on blue-collar jobs, but the occupational structure has already begun to diverge:
- Demand for AI engineers increased by 143%.
- Large tech companies have reduced entry-level hiring by 25%.
- Internship positions reduced by 30%
Future options may include either advancing to become a manager of AI agents or transitioning to industries such as manufacturing—after all, many factories also need software-savvy professionals to automate production processes.
Will there be a large-scale employment program similar to a "new policy"?
Not yet.
In July 2025, the Trump administration launched the "American AI Initiative," which includes:
- AI Education Executive Order
- Skills Training Program
- Department of Labor $84 million apprenticeship program grant
However, U.S. spending on workforce training accounts for only 0.1% of GDP, among the lowest in OECD countries. There are currently no plans to reach the scale of the WPA, which employed 8.5 million people.
Is lifelong learning worth investing in?
This is a very abstract and deeply personal question. But my answer is: yes, it’s worth it.
Inflation
If AI is truly deflationary, how would we first see signs of this?
The best indicator may be the AI API price.
GPT-4 level reasoning cost:
End of 2022: $20 per million tokens
December 2025: $0.40
Down 50 times over three years—this rate even exceeds the decline in PC computing power costs or internet bandwidth costs. It is likely to become a leading indicator of service price deflation.
How should deflation be understood if demand for knowledge products continues to grow while production costs decline?
Although AI API prices have plummeted, AI companies' revenues are surging. Lower prices → explosive usage growth → increased total spending. Meanwhile, SaaS companies are adding a 20%–37% “AI tax” during renewals. As a result, even as software production costs approach zero, SaaS revenues continue to grow.
This is similar to the computing industry during the era of Moore’s Law: individual products became cheaper, while the overall market size continued to expand.
Geopolitics
Is interconnect really important?
Extremely important.
In large GPU clusters, 30%–50% of training time is spent on communication between GPUs rather than computation.
For example, Google’s TPUv7 Ironwood connects 9,216 chips using a 3D torus topology, while Nvidia’s NVL72 connects 72 GPUs—making interconnect networks essential for scaling AI.
If a country has more energy, can it achieve AGI using outdated manufacturing processes?
It doesn't seem likely at this point.
All leading AI chips use 4nm or 3nm processes: Nvidia Blackwell, Google TPUv7, AWS Trainium3.
China's Huawei Ascend 910C (SMIC 7nm) is competitive in inference but requires more chips and more energy for training. Simply compensating for the technological gap by increasing energy consumption will eventually hit economic cost limits.
What is the most likely "Taiwan incident"?
The most likely scenario is a blockade of the Taiwan Strait.
And tensions have already been escalating:
- 2024: China conducts the "United Sword-2024B" exercise
- 2025: "Mission Justice 2025" deploys over 100 aircraft and 13 warships.
- 27 rockets were launched from Fujian, with 10 landing in Taiwan's contiguous zone.
At the same time, China has begun to distinguish between "peaceful reunification" and "reunification" in its 2026–2030 five-year plan.
TSMC is also preparing in advance: eight wafer fabs are under construction in Arizona, potentially accounting for 30% of advanced chip production in the future.
But the entire system remains on an extremely fragile balance.