Original Title: The Remarkable AGI Trades of Daniel Gross
Original Author: @johncoogan
Translation: Peggy, BlockBeats
Editor's Note: At the beginning of 2024, AI was still in a phase of enthusiasm 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 questions themselves are more inspiring than any specific predictions. The gains from AI have indeed concentrated at the infrastructure level—Nvidia has become the biggest winner; energy and electricity have rapidly become new strategic bottlenecks; API costs have plummeted while computing power, capital, and geopolitical risks have continued to intensify.
This article reviews the key questions Gross posed at the time and examines them one by one in conjunction with the real developments of the past two years. This is not only a retrospective on the logic of AI investment but also a roadmap to observe how technological revolutions reshape market structures, industrial chains, and global power dynamics.
The following is the original text:
In January 2024, Daniel Gross, then CEO of Safe Superintelligence and now head of AI products at Meta, published an article titled "AGI Trades."
The article was only one page long, listing a series of questions regarding the potential impacts of AI advancements. Looking back more than two years later, these questions appear remarkably forward-looking, even though no clear conclusions were provided at that time. Below, we review the 18 questions he posed one by one.
Markets
Where will value flow in the post-AGI world?
Currently, value is indeed concentrated at the infrastructure level—fields such as chips, packaging, and electricity. Nvidia has captured over 100% of the profits in the AI boom, as many companies are still losing money. This is also reflected clearly in market capitalization changes: Nvidia's market value increased by $3.2 trillion, from $1.2 trillion to $4.4 trillion; in contrast, the rise in cloud platforms has been much more modest (Microsoft up 4%, Amazon up 30%).
In the private market, the valuations of OpenAI, Anthropic, and xAI have also seen astonishing growth, but the combined total value increase of $1.4 trillion still falls short of Nvidia's increase in market value during the same period.
This is a crucial question right from the start of 2024.
What will happen to Nvidia and Microsoft?
Nvidia performed extremely strongly. Its revenue grew from $60.9 billion in the 2024 fiscal year to $215.9 billion in the 2026 fiscal year, nearly tripling.
Microsoft does not hold such an advantage. Azure's growth indeed accelerated to a 40% year-over-year rate, but from January 2024 to March 2026, Microsoft’s stock price rose only 4%. The market has questioned its annual AI capital expenditures exceeding $80 billion—when the investments will convert into returns remains unclear.
In this "selling shovels and picks" AI gold rush, Nvidia is clearly the biggest winner, while Microsoft's bet on infrastructure has yet to yield significant returns for shareholders.
Is copper mispriced?
Copper is indeed severely underestimated. In January 2024, the price of copper was $3.75 per pound, reaching a historical high of $6.61 per pound two years later.
The demand for copper from AI is enormous. For example:
Nvidia’s GB200 NVL72 server rack uses over 5,000 copper wires
If fully straightened, the total length exceeds 2 miles
A 100MW data center requires about 3,000 tons of copper
Overall, data centers could consume 500,000 tons of copper annually. Some have said, "Copper is the new oil." Of course, many other things have also been called "the new oil," as building AI infrastructure is extremely complex, with bottlenecks at almost every stage. Therefore, this statement should be viewed with caution.
Real Estate
If AI can write all software, will San Francisco become the new Detroit?
It depends on what "the new Detroit" refers to.
AI has actually saved San Francisco, preventing it from declining like Detroit. San Francisco is still thriving:
Office vacancy rates fell from 36.9% to 33.5%
OpenAI has 1 million square feet of office space
Anthropic has a 25-story office building
Sierra signed a lease for 300,000 square feet of office space
In the first half of 2025, 78% of U.S. AI venture capital flows into the Bay Area
Of course, there is another side: the overall employment in San Francisco remains below pre-pandemic levels, but housing prices remain strong. Therefore, it cannot be called a "ghost town" at all. The urban environment has also become cleaner.
How will AI impact wealth inequality?
It is still too early to draw conclusions, as data changes are not yet significant, but some studies are worth noting.
The IMF's 2025 research suggests that AI may reduce wage inequality (due to automation of high-income jobs), but it may exacerbate wealth inequality (capital gains concentrated in the hands of tech company owners).
OECD research has found that wages for low-skilled jobs are growing fastest (assemblies +11.6%), while those for high-skilled jobs are growing slowest (CEO +2.7%). However, this may more reflect minimum wage policies rather than AI itself.
In the capital markets, concentration is also rising: the "Magnificent Seven" (Mag7) account for about 32% of the S&P 500 market capitalization and contributed about 42% of total returns in 2025; at the same time, the massive financing of AI startups (OpenAI $110 billion, Anthropic $30 billion) has also bestowed massive private wealth on a few founders and investors.
Energy & Data Centers
If AI turns into an energy competition, how should one invest?
This judgment is completely correct. AI has indeed turned into an energy game.
Those who grasped this transaction made a significant profit. For example:
Vistra: +321%, the second largest increase in the S&P for 2024 (only behind Palantir)
Constellation Energy: tripled since the release of ChatGPT
NRG Energy: approximately 95% increase in 2025 alone
Oklo: over 700% increase in 12 months
Nuclear energy has experienced an explosion:
Microsoft signed a $16 billion, 20-year PPA, restarting the Three Mile Island nuclear power plant
Google signed a 500MW small modular reactor (SMR) agreement with Kairos Power
Meta signed a 6.6GW power contract with multiple nuclear energy companies
Energy has become one of the most successful investment themes of the AI era.
In the entire data center supply chain, which links are the hardest to expand 10 times?
The bottleneck in the chip industry is CoWoS packaging technology (TSMC's Chip-on-Wafer-on-Substrate).
In the data center field, the biggest bottleneck may be power transformers.
Delivery cycles approach 3 years
A 30% supply gap is expected in 2025
Costs have risen 150% since 2020
This century-old technology has become a critical limit on the speed at which data centers can connect to the grid.
Is coal undervalued?
To some extent, yes, but far less than copper. In 2025, coal prices actually dropped by about 22%, rebounding slightly by early 2026.
Coal companies performed reasonably well:
Peabody Energy: +34%
CONSOL Energy: +37%
Meanwhile, U.S. coal electricity generation is expected to grow by 13% by September 2025.
States with rapid data center growth have shown particularly pronounced performances:
Ohio: +23%
Oklahoma: +58%
Nations
Who are the winners and who are the losers?
The clear winner is the United States.
Private AI investment in the U.S. reached $109 billion in 2024 (only $9.3 billion in China)
Cumulative investment since 2013 exceeds $470 billion, more than all other countries combined
The U.S. released 40 significant AI models in 2024, while China released 15
The game is not over, but for now, the U.S. is the center of AI competition.
What will happen to India's $250 billion GDP reliant on GPT-4 tokens?
The situation is beginning to emerge but is still in the early stages. Hiring in India's IT outsourcing industry has notably declined. Between 2024 and 2025, large IT companies are cutting about 58,000 jobs, whereas between 2021 and 2023, the industry added 360,000 employees.
Will software engineers be replaced like typists were in history?
At present, software engineers have not become blue-collar workers, but there is already a differentiation in the job structure:
Demand for AI engineers increased by 143%
Large tech companies cut entry-level recruitment by 25%
Internship positions decreased by 30%
Future choices may be: either upgrade to "AI agent managers," or shift towards manufacturing and other areas—after all, many factories also need software-savvy individuals to automate production processes.
Will there be a large-scale employment plan similar to the "New Deal"?
Not currently.
In July 2025, the Trump administration launched the "American AI Action Plan," consisting of:
AI education executive order
Skills training programs
$84 million apprenticeship grant from the Department of Labor
However, U.S. workforce training expenditure accounts for only 0.1% of GDP, one of the lowest among OECD countries. There are currently no plans to reach the scale of the WPA of that year (8.5 million employment program).
Is lifelong learning worth investing in?
This is a very abstract and personal question. But my answer is: Yes.
Inflation
If AI truly has deflationary properties, how will we first see this signal?
The best indicator may be the prices of AI APIs.
Cost of reasoning at a GPT-4 level:
End of 2022: $20 per million tokens
December 2025: $0.40
A decrease of 50 times in three years. This rate of decline even surpasses that of the decrease in PC computational costs or internet bandwidth costs. This is likely to become a leading indicator of service price deflation.
If demand for knowledge products continues to grow, while production costs decline, how should we understand deflation?
Although AI API prices have plummeted, AI companies' revenues are soaring. Price decline → explosive usage → total expenditure increases. Meanwhile, SaaS companies are also adding a "20%-37% AI tax" at renewal. Therefore, even if the cost of software production approaches zero, SaaS revenues are still growing.
This is similar to the computing industry during the era of Moore's Law: individual products become cheaper, but the overall market scale continues to expand.
Geopolitics
Is interconnection 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 TPUv7 Ironwood connects 9,216 chips using a 3D torus topology
Nvidia NVL72 connects 72 GPUs
Therefore, interconnection networks are crucial for AI scalability.
If a country has more energy, can it achieve AGI with lagging processes?
Currently, this seems unlikely.
All leading AI chips are using 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 energy in training. Simply increasing energy consumption to bridge the technological gap will ultimately encounter limits of economic cost.
What is the most likely "Taiwan incident"?
The most likely scenario is a blockade of the Taiwan Strait.
And tensions are escalating:
2024: China held the "Joint Sword-2024B" exercise
2025: "Mission of Justice 2025" mobilized over 100 aircraft and 13 warships
27 rockets were launched from Fujian, 10 of which landed in Taiwan's contiguous zone
At the same time, China has started to separate "peaceful unification" and "unification" in its 2026-2030 five-year plan.
Taiwan Semiconductor Manufacturing Company is also laying groundwork: Arizona is building eight wafer fabs, which may take on 30% of advanced chip capacity in the future.
However, the entire system remains on an extremely fragile balance.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
