AI earnings: Big growth, bigger spending, tougher questions

Key points

• AI demand remains strong, but the cost of growth is rising. Cloud, semiconductor and memory companies continue to report exceptional growth. However, soaring capital expenditure is shifting investor attention from revenue growth to cash flow, profitability and returns on investment.
• AI infrastructure is still booming, but exceptional margins may be harder to sustain. Supply constraints are supporting strong earnings across chips, memory and storage. However, elevated profitability encourages capacity expansion, potentially increasing competition and pricing pressure over time.
• The next phase of AI is adoption, but the productivity payoff remains largely unproven. Early monetisation is visible in advertising, cloud and enterprise software, but broad-based productivity gains have yet to emerge. For portfolios, this argues for financial resilience, valuation discipline and diversification beyond AI infrastructure into potential beneficiaries of adoption.

The latest round of technology earnings has delivered a clear message: AI demand is not the problem. The challenge is increasingly about the cost of meeting that demand and how much of the growth eventually translates into profits and cash flow.

Cloud providers are growing rapidly, semiconductor and memory companies are enjoying exceptional demand, and AI is beginning to generate revenues for software businesses.

Yet strong results have not always been rewarded by investors. Higher spending, uncertain future margins and elevated valuations are making markets more selective.

Here is what the latest results tell us about where the AI investment cycle stands.

Tech earnings: What the major companies reported

Big Tech: Strong growth, rising investment commitments

• Microsoft: Revenue grew 19% year-on-year, with Azure revenue up 48%. The company generated $18.6 billion in quarterly free cash flow despite substantial infrastructure spending. Microsoft continues to demonstrate the benefits of combining cloud growth with an established enterprise customer base, although AI investment requirements remain significant.
• Alphabet (Google): Revenue rose 25%, with Google Cloud revenue more than doubling to $30.7 billion. Search revenue grew 12%, showing resilience in its core advertising business. However, substantial capital spending pushed quarterly free cash flow into negative territory, reinforcing questions about the cost of AI expansion and its longer-term impact on Search economics.
• Amazon: AWS revenue grew 39%, its fastest pace in nearly five years, while Amazon delivered a record operating margin of 13.9%. Yet expected 2026 capital spending of approximately $220 billion shows how expensive the cloud expansion has become. AWS growth is encouraging, but the returns on this investment will remain a key focus.
• Meta: Revenue increased 26%, supported by 10% growth in advertising impressions and a 14% increase in average ad prices. AI is already improving its advertising business. However, quarterly capital spending reached nearly $35 billion, while free cash flow turned negative. Reported profitability was also affected by a substantial one-off litigation charge. The question is whether future AI benefits can justify increasingly large investments.

Semiconductors, memory and storage: Strong demand and exceptional profitability

• TSMC: Revenue increased 41% in US-dollar terms, with a gross margin of 66.4%. Demand for advanced chips remains strong, reinforcing TSMC's importance in the AI supply chain. However, the strength of current margins also raises the bar for future earnings.
• ASML: Quarterly revenue reached €11.7 billion, with a gross margin of 57.5%. Demand for advanced manufacturing equipment and productivity upgrades continues to improve, with customers discussing capacity needs extending several years into the future. This provides visibility, although the industry remains exposed to capital-spending cycles.
• SK hynix: Revenue surged 328% year-on-year, while operating margin reached 80%, supported by strong AI memory demand and pricing. Long-term supply agreements provide greater visibility, but such exceptional profitability also raises questions about how margins might evolve as industry capacity expands.
• Seagate: Revenue reached $4.2 billion and quarterly free cash flow increased to $1.5 billion, supported by strong demand for high-capacity storage. The results show how AI infrastructure spending is benefiting businesses beyond chipmakers. However, storage remains a cyclical industry, and strong current profitability may attract additional supply.

Enterprise technology: Growth opportunities, but not without disruption

• ServiceNow: Subscription revenue increased 21%, while the company raised its 2026 AI annual contract value target to $1.7 billion. This provides evidence of enterprise customers paying for AI capabilities. However, concerns about forward contracted-revenue growth and the potential disruption of traditional software models weighed on investor sentiment.
• Intel: Revenue increased 21%, with data-centre and AI revenue growing 78%. However, a weaker outlook for the fourth quarter and continued weakness in the PC market highlighted the uneven recovery across its businesses. Strong AI demand alone is not enough to offset every operational challenge.

Four key themes from the earnings season

1: AI demand is strong, but the spending bill is getting much bigger

The clearest positive takeaway is that demand for AI infrastructure continues to accelerate.

Azure grew 48%, Google Cloud expanded 103%, and AWS grew 39%. These numbers suggest that the appetite for computing capacity remains substantial.

However, meeting that demand is becoming increasingly expensive.

Amazon expects to spend approximately $220 billion on capital expenditure this year, while Microsoft, Alphabet and Meta are also committing substantial resources to expanding data centres and AI infrastructure.

This spending supports near-term growth across the semiconductor supply chain, but it creates longer-term financial obligations.

Depreciation, equipment replacement, energy costs and financing requirements will increasingly matter. AI technology is also evolving rapidly, meaning infrastructure purchased today could become less competitive sooner than expected.

The risk is not necessarily that AI demand disappoints. It is that the cost of supplying that demand rises faster than the profits it generates.

For investors, the distinction between companies capable of financing investment through recurring cash flow and those reliant on external capital is becoming more important.

2: AI benefits are emerging, but monetisation remains uneven

The earnings show that AI is starting to deliver commercial benefits, but not all companies are benefiting equally.

Meta is already seeing improvements in advertising performance. Microsoft is monetising AI through cloud infrastructure and enterprise software. ServiceNow is reporting growing AI-related contract value.

These are encouraging signs.

However, higher AI revenues do not automatically translate into stronger profits. Some companies are generating incremental revenues from AI, while others are primarily absorbing higher infrastructure and development costs.

Traditional software businesses face an additional complication. AI could improve their products and create new revenue streams, but AI agents may also reduce demand for some existing software functions or challenge per-user subscription pricing.

This is creating a more complex environment for investors.

The next question is not simply whether customers are using AI, but whether companies can capture enough economic value from that usage.

3: AI infrastructure remains strong, but exceptional margins may become harder to sustain

Some of the most impressive earnings this season came from companies supplying AI infrastructure.

TSMC reported gross margins above 66%, SK hynix achieved an 80% operating margin, and advanced semiconductor equipment and storage companies continue to benefit from strong demand.

These results reflect genuine supply constraints and technological advantages.

However, unusually high profitability also creates incentives for the industry to expand capacity.

Over time, new manufacturing facilities, additional memory production and greater availability of computing infrastructure could reduce shortages and put pressure on pricing.

That does not mean a downturn is imminent. Advanced semiconductor manufacturing is difficult to replicate, and supply constraints could persist longer than expected, particularly in highly specialised areas.

But the investment equation is changing.

When earnings expectations already assume exceptional growth and margins, companies need increasingly strong results to deliver positive surprises.

The infrastructure boom can continue while share-price returns become more uneven. Strong demand does not guarantee that today's exceptional margins will last indefinitely.

4: The next phase of AI is adoption — but the productivity payoff is still largely unproven

So far, much of the AI investment cycle has focused on building capacity: chips, data centres, networking, memory and power infrastructure.

The next phase should increasingly focus on what businesses can actually achieve with that infrastructure.

AI has the potential to improve productivity across industries, from automating administrative work and customer service to accelerating research, software development and business processes.

There are already examples of benefits in individual businesses. Meta's advertising improvements and growing enterprise AI software sales suggest that commercial use cases are developing.

However, there is still limited evidence of broad, measurable productivity gains across the wider economy.

Companies may be adopting AI tools, but implementation costs, employee training, workflow changes and the need for human oversight can delay financial benefits.

Importantly, productivity gains do not always accrue to the company purchasing the AI tools. Some benefits may be passed on to customers through lower prices or absorbed by competition.

This could be one of the most important questions for 2027.

If adoption begins to generate sustained improvements in corporate margins and returns, the AI investment opportunity could broaden beyond technology infrastructure into a much wider range of sectors.

If the productivity gains take longer to materialise, businesses may become more cautious about AI spending, potentially slowing the growth of the infrastructure suppliers that have benefited most so far.

The next stage of the AI cycle may be defined less by how much infrastructure gets built and more by how effectively businesses put it to work.

Beyond the earnings: Where might investors be overreacting?

One of the more interesting developments this earnings season has been the difference between business performance and share-price reactions.

Some companies have experienced sharp selloffs despite relatively resilient underlying businesses. Others are facing challenges that may justify a more lasting reassessment of their valuations.

There are three areas worth distinguishing.

Bottom line: The AI story is entering a more demanding phase

The latest earnings have reinforced confidence in the strength of AI demand. Cloud businesses are growing rapidly, infrastructure suppliers are reporting exceptional results, and commercial applications are beginning to generate revenue.

But the investment story is becoming more complicated.

Spending is rising, infrastructure margins may not remain exceptional forever, and the broader productivity benefits of AI are still largely ahead of us.

For investors, the next phase calls for a more balanced approach: maintain awareness of the structural AI opportunity, pay closer attention to cash flow and returns on capital, and look for evidence that AI adoption is creating sustainable economic value beyond the technology sector.

The AI boom has already shown us how much companies are willing to spend. The next chapter will be about how much they can earn from that investment.

Read the original analysis here