Who Took the Money in the AI Era? A Must-See Investment Checklist for HALO Asset Trading

Original Title: The Remarkable AGI Trades of Daniel Gross
Original Author: @johncoogan
Translation: Peggy, BlockBeats

Editor's Note: In early 2024, AI was still in a stage of simultaneous frenzy and uncertainty. At that time, Daniel Gross presented 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, more than any specific predictions, were more inspiring. The benefits of AI did indeed concentrate at the infrastructure layer—NVIDIA emerged as the biggest winner; energy and power quickly became new strategic bottlenecks; API costs plummeted, while computing power, capital, and geopolitical risks continued to escalate.

This article reviews the key questions Gross raised at the time and examines them one by one in light of the reality of the past two years. This is not only a retrospective of AI investment logic but also a roadmap observing how the technological revolution is reshaping market structures, industrial chains, and global power dynamics.

The following is the original text:

In January 2024, Daniel Gross, who was then the CEO of Safe Superintelligence and is now the Head of Product at Meta AI, published an article titled "AGI Trades."

This article was only one page long, listing a series of questions about the potential impacts of AI progress. Looking back more than two years later, these questions appear particularly forward-thinking, even though each question did not provide a clear conclusion at the time. Below, we will review the 18 questions he posed one by one.

Markets

In a post-AGI world, where will value flow?

Currently, value has indeed concentrated at the infrastructure layer—chips, packaging, power, and other areas. NVIDIA has nearly taken away over 100% of the profits in the AI boom, as many companies are still operating at a loss. This is also very evident in the changes in market value: NVIDIA's market cap has increased by $3.2 trillion, rising from $1.2 trillion to $4.4 trillion; in comparison, the growth of cloud platforms has been much more modest (Microsoft up 4%, Amazon up 30%).

In the private market, the valuation growth of OpenAI, Anthropic, and xAI has also been remarkable, but the total value increase of the three, totaling $1.4 trillion, is still lower than the market cap increase NVIDIA achieved during the same period.

This is a very critical question right from the start of 2024.

What will happen to NVIDIA and Microsoft?

NVIDIA has shown extremely strong performance. Its revenue has grown from $60.9 billion in FY2024 to $215.9 billion in FY2026, almost tripling.

Microsoft, on the other hand, is not as dominant. While Azure's growth has indeed accelerated to a 40% year-over-year growth rate, Microsoft's stock price has only risen by 4% from January 2024 to March 2026. The market has questioned its annual AI capital expenditure of over $80 billion — when the investment will translate into returns remains unclear.

In this "selling shovels in a gold rush" AI frenzy, NVIDIA is clearly the biggest winner, while Microsoft's bet on infrastructure has yet to deliver significant returns to shareholders.

Is Copper Mispriced?

Indeed, it has been 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 in AI is enormous. For example, the NVIDIA GB200 NVL72 server rack uses over 5000 copper wires. If all stretched out, the total length would exceed 2 miles, and a 100MW data center would require about 3000 tons of copper.

Overall, data centers may consume 500,000 tons of copper per year. Some say "Copper is the new oil" because of this. Of course, many other things are also called the "new oil" as AI infrastructure deployment is highly complex with bottlenecks at nearly every level. So, this statement should also be viewed with caution.

Real Estate

If AI can write all software, will San Francisco become the new Detroit?

That depends on what is meant by the "new Detroit."

AI has actually saved San Francisco from becoming a city in decline like Detroit. San Francisco continues to thrive:

· Office vacancy rates have decreased from 36.9% to 33.5%

· OpenAI owns 1 million square feet of office space

·Anthropic owns a 25-story office building

·Sierra signs 300,000 square feet of office space

In the first half of 2025, 78% of U.S. AI venture capital flows to the Bay Area. Of course, there is another side: the overall employment in San Francisco is still below pre-pandemic levels, but real estate prices remain strong. So, it is by no means a "hollow city." The urban environment has also become cleaner.

How Will AI Impact Wealth Inequality?

It is still too early to draw conclusions, as the data changes are not clear, but there are some studies worth noting.

The IMF's 2025 study suggests that AI may reduce wage inequality (due to automation of high-income jobs) but could exacerbate wealth inequality (capital gains concentrated in the hands of tech company owners). OECD's research found that wage growth is fastest for low-skilled jobs (assembly workers +11.6%) and slowest for high-skilled jobs (CEOs +2.7%), but this may reflect more on minimum wage policies rather than AI itself.

In the capital markets, concentration is also increasing: the "Big Seven" (Mag7) account for around 32% of the S&P 500 market cap, contributing approximately 42% of the total return in 2025; at the same time, massive funding for AI startups (OpenAI $1.1T, Anthropic $300B) has also allowed a few founders and investors to amass huge personal wealth.

Energy & Data Centers

If AI Becomes an Energy Race, How Should You Invest?

This assessment is absolutely correct. AI has indeed become an energy game.

Those who seized this opportunity have made a lot of money. For example:

·Vistra: +321%, the second-largest gain in the S&P 500 in 2024 (second only to Palantir)

·Constellation Energy: tripled in stock price since the release of ChatGPT

·NRG Energy: nearly 95% increase in a single year in 2025

·Oklo: over 700% increase in 12 months

The Nuclear Energy Boom:

· Microsoft signs $16 billion, 20-year PPA, restarting Three Mile Island nuclear power plant

· Google signs 500MW Small Modular Reactor (SMR) agreement with Kairos Power

· Meta secures 6.6GW power contracts with multiple nuclear energy companies

Energy has become one of the most successful investment themes in the AI era.

Within the entire data center supply chain, which links are most challenging to scale by 10x?

The bottleneck in the chip industry is CoWoS packaging technology (TSMC's Chip-on-Wafer-on-Substrate).

In the data center space, the biggest bottleneck may be power transformers.

· Lead times approaching 3 years

· 30% supply gap expected by 2025

· Costs have risen 150% since 2020

This century-old technology has become a key constraint on the speed at which data centers can connect to the grid.

Is coal being underestimated?

To some extent, yes, but far less so than copper. Coal prices are actually expected to drop by around 22% by 2025 and recover slightly by early 2026.

Coal companies have performed reasonably well:

· Peabody Energy: +34%

· CONSOL Energy: +37%

Meanwhile, U.S. coal-fired electricity generation has increased by 13% through September 2025.

States with rapidly growing data centers show particularly strong performance:

· Ohio: +23%

· Oklahoma: +58%

Nations

Who are the winners and losers?

The winner is clearly the United States.

In 2024, US private AI investment reached $109 billion (China only $9.3 billion), totaling $470 billion since 2013, surpassing the sum of other countries. In 2024, the US released 40 major AI models, China 15.

The game is not over yet, but as it stands, the US is at the center of the AI competition.

What will happen to India's $250 billion GDP export reliance on the GPT-4 token?

The situation is beginning to unfold, but still in its early stages. Recruitment in the Indian IT outsourcing industry has significantly declined. Between 2024 and 2025, large IT companies will lay off about 58,000 people, whereas between 2021 and 2023, the industry had added 360,000 employees.

Will software engineers be replaced like typists in history?

Currently, software engineers have not resorted to blue-collar work, but the professional structure has already shown differentiation:

· AI Engineer demand increased by 143%

· Large tech companies decreased junior position hiring by 25%

· Internship positions reduced by 30%

The future choices may be: either move up to become an "AI agent manager" or pivot to fields like manufacturing—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 at the moment.

In July 2025, the Trump administration launched the "US AI Action Plan," including:

· AI Education Executive Order

· Skills training programs

· $84 million Department of Labor apprenticeship program grants

However, US labor training expenditure accounts for only 0.1% of GDP, nearly the lowest among OECD countries. Currently, there is no plan to reach the scale of the WPA (8.5 million employment plan) of that year.

Is lifelong learning worth investing in?

This is a very abstract and personal question. But my answer is: yes, it is worth it.

Inflation

If AI is truly deflationary, how might we first see that signal?

The best indicator might be the price of AI APIs.

Cost of reasoning at GPT-4 level:

· End of 2022: $20 per million tokens

· December 2025: $0.40

A 50x decrease over three years. This rate is even faster than the decline in PC compute costs or internet bandwidth costs. This is likely to be a leading indicator of price deflation for services.

If demand for knowledge products continues to rise while production costs fall, how should deflation be understood?

While the price of AI APIs plummets, AI company revenues are soaring. Price drop → explosion in usage → total spending increases. Meanwhile, SaaS companies are adding a 20%–37% "AI tax" on renewals. So, even as software production costs approach zero, SaaS revenue is still growing.

This is akin to the computing industry in the era of Moore's Law: individual products get cheaper, but the overall market size continues to expand.

Geopolitics

Is interconnect truly important?

Extremely important.

In large GPU clusters, 30%–50% of training time is spent on GPU-to-GPU communication rather than computation.

For example, Google TPUv7 Ironwood uses a 3D torus topology to connect 9216 chips, and Nvidia NVL72 connects 72 GPUs, making the interconnect crucial for AI scalability.

If a country has more energy, could it achieve AGI with outdated processes?

Currently, it seems unlikely.

All leading AI chips are on a 4nm or 3nm process, such as Nvidia Blackwell, Google TPUv7, AWS Trainium3

China's Huawei Ascend 910C (SMIC 7nm) is competitive in inference but requires more chips and more power for training. Simply bridging the technical gap by increasing energy consumption will eventually run into economic cost constraints.

What is the most likely "Taiwan Event"?

The most likely scenario is a blockade of the Taiwan Strait.

The tension has been escalating:

· 2024: China holds the "Joint Sword-2024B" exercise

· 2025: "Just Mission 2025" mobilizes over 100 aircraft and 13 warships

· 27 rockets launched from Fujian, with 10 landing in Taiwan's contiguous zone

Meanwhile, in its 2026–2030 five-year plan, China begins to differentiate between "peaceful reunification" and "unification."

TSMC is also proactively preparing: building 8 fabs in Arizona, potentially accounting for 30% of advanced chip capacity in the future.

Yet, the entire system remains on an extremely fragile equilibrium.

[Original Article Link]