by Joe Herrle, CFA
Senior Vice President
Alternative Assets and Portfolio Management
The numbers coming out of the first quarter earnings season were, by any historical standard, staggering. Amazon, Alphabet, Microsoft and Meta, the four largest cloud and technology platforms, accounted for $650 to $700 billion in capital expenditures (capex) for 2026. That is nearly double what the same group spent in 2025, and larger than the GDP of most countries. The AI infrastructure buildout is no longer a promise; it is a fact inscribed in balance sheets.
Source: Company earnings reports and guidance, Q1 2026. Microsoft figure is FY2026 estimate. All figures in USD billions.
Amazon alone projects $200 billion in capex this year, a figure CEO Andy Jassy defended in April’s annual shareholder letter with characteristic conviction: "We're not going to be conservative in how we play this." Alphabet raised its 2026 guidance to $180–190 billion after spending $35.7 billion in a single quarter, with CFO Anat Ashkenazi flagging that 2027 capex will "significantly increase" from there. Meta lifted its range to $125–145 billion, a figure large enough that its disclosure triggered a 6% after-hours selloff — the market's way of asking a simple question: who exactly captures the return on all this spending?
That question is the right one for investors to be asking right now.
The spend is real. The revenue conversion is uneven.
To be fair, some evidence of return is emerging. Google Cloud revenue surged 63% year-over-year to $20 billion in the first quarter of 2026, topping Wall Street estimates by nearly $2 billion, the clearest single data point that infrastructure investment is generating customer demand. Microsoft's AI business now runs at an annualized revenue rate of $37 billion, up 123% year-over-year, while its Azure cloud service grew 40%. Amazon Web Services (AWS) posted its fastest growth rate in 15 quarters at 28%.
These are exceptional numbers, but they also illustrate a divergence: Google Cloud is now outspending Microsoft on capex despite running a cloud business roughly 37% its size. The smaller player is spending aggressively to close a gap. Meanwhile, Amazon's free cash flow collapsed to $1.2 billion as a $59.3 billion year-over-year surge in infrastructure spending consumed nearly all of its operating cash flow. The math of AI investment is, in many cases, front-loaded cost and back-loaded revenue.
The cost curve is moving the wrong way.
Compounding the capex surge is the inflation embedded within it. The International Data Corporation forecasts that DRAM, a type of memory chip critical to desktop and mobile devices, will cost $9.71 per gigabyte in 2026, compared to $3.76 in 2025. Spot prices for Nvidia H200 GPUs have risen sharply. Microsoft attributed roughly $25 billion of its 2026 capex to component price inflation, a figure that underscores how much of the spending increase is cost-driven, not purely demand-driven, capacity expansion. The hyperscalers are absorbing these increases for now. Whether they can pass them through to enterprise customers over time is a key variable.
So, who actually benefits?
This is the portfolio question that matters. At Ferguson Wellman, we think about the AI infrastructure cycle in three distinct layers, each with a different risk-reward profile:
1) Builders: The Hyperscalers themselves. They are generating real cloud revenue growth and demonstrating that enterprise AI demand is durable. However, capex-to-revenue ratios have reached historically elevated levels: Meta is tracking toward capex equal to roughly 54% of sales, Microsoft 47% and Alphabet 46% in 2026. These ratios are not indefinitely sustainable and create meaningful depreciation and financing headwinds in future years, particularly as Alphabet issued a rare 100-year "century bond", and Amazon raised roughly $54 billion in March to help finance its buildouts.
2) Enablers: Semiconductor and hardware companies supplying infrastructure, such as GPU manufacturers, custom silicon designers, memory producers and networking equipment providers. Two weeks ago, our colleague Peter Jones, CFA, noted in his blog "Silicon Surging" that this group produced the most explosive earnings of the quarter and currently sits at the largest weight in the S&P 500 of any sector. The risk here is cyclicality; upcycles of this magnitude historically carry the seeds of overcapacity.
But the enabler category extends well beyond silicon. Regulated and independent power producers have emerged as critical infrastructure counterparts to the data center boom: hyperscale facilities are extraordinarily power-dense, and the grid buildout required to serve them is generating a multi-year capital deployment cycle for utilities with favorable load growth exposure. Industrial companies — electrical equipment manufacturers, cooling systems providers, steel and structural fabricators, and engineering and construction firms — are similarly capacity-constrained as data center construction backlogs extend into 2027 and beyond.
3) Applications: Where the long-term value creation likely accrues, but where the investment case is least defined today. Which software companies successfully integrate AI to meaningfully expand margins or total addressable markets, and which are disrupted by it, remains the central unresolved question of this investment cycle. For further reading, Krystal Daibes Higgins, CFA, recently addressed software selloff in earlier blogs, including "The Cost of Perfection" and "SaaSination or Selective Opportunity?"
Our takeaway
The AI supercycle is real, and the infrastructure investment underlying it is the largest, most concentrated capital deployment in technology history. But the scale of spending does not automatically translate to proportional investment returns. We remain constructive on AI beneficiaries across all three layers, while paying close attention to free cash flow generation, capex payback periods and the emerging divergence in cloud growth rates as leading indicators of where durable economic value is actually accruing. The companies that own both the model layer and the infrastructure (and can convert that advantage into pricing power) appear best positioned for the next phase of this cycle.
The magnificent capex is being spent. The question now is who earns it back.
Takeaways for the Week
April nonfarm payrolls came in at 115,000, well above the consensus estimate of 65,000, while the unemployment rate held steady at 4.3%. The print reinforced the resilience of the labor market despite elevated energy costs and geopolitical headwinds, and further diminished expectations for Fed rate cuts in 2026. Fed funds futures currently suggest the odds of rates remaining unchanged for the remainder of the year at roughly two in three
The S&P 500 closed at a new record on Wednesday, advancing 1.46%, as reports that the U.S. and Iran were nearing a potential peace agreement. On this news, oil prices fell sharply, lifting equities broadly. The ceasefire narrative reversed on Thursday after Iran alleged U.S. violations, illustrating how headline-driven this market remains. Despite the back-and-forth, the S&P 500 is on track for its sixth consecutive winning week
S&P 500 earnings growth is tracking near 28% year-over-year, with the percentage of companies beating earnings estimates above long-term averages, a strong result that has kept equity momentum intact even as macro crosscurrents persist
Disclosure: The views expressed represent the opinion of Ferguson Wellman. The views are subject to change and are not intended as a forecast or guarantee of future results. This material is for informational purposes only. It does not constitute investment advice and is not intended as an endorsement of any specific investment. Statements of future expectations, estimates, projections and other forward-looking statements are based on available information and Ferguson Wellman’s views as of the time of these statements. Past performance may not be indicative of future results. Ferguson Wellman, Octavia Group and West Bearing do not provide tax, legal, insurance or medical advice. This material has been prepared for general educational purposes only and not as a substitute for qualified counsel who can determine how this information applies to you. We believe the information provided is from reliable sources but should not be assumed accurate or complete.
Please see additional disclosures.