Capital: numbers that don’t fit into conventional frameworks
The numbers speak for themselves. Goldman Sachs analysts predict that the four largest players in this field – Meta, Microsoft, Amazon, and Alphabet – will have collectively spent approximately $5.3 trillion on investments in this area from 2025 to 2030. A even larger estimate was recently announced – $7.6 trillion from 2026 to 2031, which included data processing infrastructure, data centers, and energy. For comparison, this is more than 90 times Lithuania’s GDP.
Read more UN urges AI companies to declare their environmental impact
This scale of funding necessitates unconventional solutions. Public capital markets alone cannot bear such a weight. Financial market analysts emphasize that in the coming years, a combination of various financing structures will be a necessity – not an option. Private capital market representatives are already actively participating in this process: in 2025, capital raised by infrastructure funds reached a new annual record of $221 billion, with an annual return exceeding 12%, outperforming all other private capital categories except private venture capital funds.
Infrastructure as a long-term catalyst
Infrastructure and related tools in this context represent perhaps the most attractive long-term opportunity. From 2021 to 2024, this market grew by approximately 12% per year. Analysts see the potential for this pace to return to even higher results, possibly reaching 16-17% annual growth, which prevailed from 2012 to 2021. At this pace, the amount of assets under management could reach $3 trillion by 2030.
The attractiveness of this asset class is strengthened by several frequently used arguments: revenue stream, inflation protection (some contracts include direct cost pass-through mechanisms), and lower correlation with classical investments, which is expected to reduce market volatility risk. For institutional investors, who have recently increased their bond holdings, this sector is likely to be attractive, especially in pursuit of higher returns and considering declining yields.
Data centers currently account for two-thirds of all digital infrastructure transaction volume. Together with utilities, this category accounted for more than half of the total infrastructure transaction value in the first quarter of 2026.
Semiconductors – structural break or cycle peak?
The semiconductor sector is currently the hottest and most debated in the entire market. The Philadelphia Stock Exchange Semiconductor Index has jumped about 70% in the last two months, and memory chip manufacturers have achieved stunning results: Micron shares have tripled this year, and Asian competitors have seen even more dramatic gains.
High-bandwidth memory chips (HBM) have fundamentally changed supply dynamics. They are more complex to manufacture, have a higher defect rate, and occupy a disproportionately large share of manufacturing capacity. This structurally limits supply and keeps prices high – unlike typical memory chip cycles where oversupply quickly erodes margins.
Micron’s revenues are projected to be $66.8 billion in 2026, compared to $8.5 billion in 2025. In 2027, approximately $120 billion is expected – more than Amazon in the same period.
However, the bearish perspective is equally significant. The semiconductor index is currently trading at 71 times earnings – the most expensive level since the end of the 2008 financial crisis. By sales ratio, it’s even 15 times, which is the highest indicator since 2002, more than three times the long-term average. Valuations are based on the assumption that the boom will continue. If demand falters even briefly – either due to macroeconomic reasons or a review of the aforementioned major companies’ investment plans – the correction could be sudden and deep.
IPO wave – new players in public markets
2026 also marks a new stage in the history of AI ecosystem financing – the largest private players are approaching listings. SpaceX plans the largest IPO in history to date – a $75 billion issuance at a valuation of $1.8 trillion. Anthropic, whose valuation already exceeds OpenAI and reaches $965 billion, has filed a confidential IPO request. OpenAI is preparing to follow suit.
Alphabet announced a plan for an $80 billion stock issuance in early June – this is likely to become the largest equity raising transaction in financial market history. The move is exceptional – Google’s parent company has not issued shares for twenty years. However, the pace of AI infrastructure construction has become so rapid that even Alphabet’s massive revenue streams can no longer finance it. The company expects capital expenditures to reach $180-190 billion in 2026 – double that of 2025, and to grow further in 2027.
These are not just financial transactions. This is a signal that segments of the financial market ecosystem, even with record recent flows, can no longer individually meet the financing needs of these companies’ growth. Companies with annual revenue growth of several hundred percent inevitably hit limits and must combine financing sources.
For investors, this creates two different scenarios. If the IPO succeeds – it opens a new stage of liquidity in the AI ecosystem and potentially further raises sector valuations. If any of these listings fail or disappoint – a domino effect could quickly spread throughout the sector, including private investments and semiconductor stocks.
Questions the market still doesn’t have answers to
The first question – is demand real or debt-financed? The four largest companies mentioned plan to spend up to $725 billion on capital in 2026, but companies are increasingly financing these expenditures with debt. This means that the continuity of investments depends not only on the profitability of AI business models but also on financing conditions. The interest rate environment here is not a secondary, but a primary risk.
The second question – where is the limit for data center construction? Data center construction costs in the US are growing by 28% per year, and in Australia, one in five dollars for non-residential construction is already allocated to data centers. However, the pipeline of planned but not yet started projects is enormous. Goldman Sachs analysts point out that the investment trajectory of the aforementioned four major companies in 2027 is a decisive factor that will determine the growth of data center infrastructure in 2028. This means that the current construction boom is partly based on expectations, not confirmed orders.
The third question – what happens after construction? Both US and global data show that the operational phase of data centers creates a more modest macroeconomic effect than the construction phase. The economic effect may be significantly smaller than sector enthusiasts promise.
Read more What to do and what to avoid before a blood test: this can fundamentally change the results
