Is Bytes Technology the Canary in the AI Coal Mine?

For most of the artificial intelligence boom, investors have focused on one question: how strong is demand?

The answer has repeatedly been extremely strong. Demand for advanced semiconductors, cloud capacity, memory, networking equipment and data centres continues to exceed expectations. Rising capital expenditure has therefore been treated as confirmation that the AI investment cycle remains intact.

A more important question is now emerging: can that demand generate an acceptable return on the extraordinary amount of capital being invested?

A recent downgrade of Bytes Technology may provide an early warning.

A warning from further down the supply chain

Bytes is one of the UK’s largest software, cloud and cybersecurity resellers, with a particularly important relationship with Microsoft.

UBS recently downgraded the company from Neutral to Sell, arguing that changes to Microsoft’s partner incentives could place renewed pressure on its profitability. Microsoft appears to be reducing the economics available from routine software renewals while offering greater rewards for selling Copilot, increasing Azure consumption and winning new customers.

Bytes is particularly exposed because Microsoft-related incentives and mark-ups are estimated to generate approximately half its gross profit.

The downgrade follows a year in which Bytes’ gross invoiced income increased by 11.5%, but gross profit rose by only 2.5% and operating profit fell by 5.6%. Previous changes to Microsoft incentives were one reason why higher customer spending failed to translate into equivalent profit growth.

Bytes is not evidence that companies have stopped buying technology. It is evidence that strong technology spending does not guarantee that every participant in the ecosystem will retain its share of the profits.

That distinction increasingly matters across the AI trade.

From revenue growth to returns on capital

The hyperscalers continue to report formidable demand. Microsoft’s Azure business grew 39% in constant currency in its most recent quarter, with demand still exceeding available capacity.

However, Microsoft’s cloud gross margin fell to 66%, with the company explicitly attributing some of that pressure to AI infrastructure investment and greater usage of AI products.

Microsoft’s forthcoming earnings are therefore important well beyond the company itself. Investors will be watching Azure growth, Copilot adoption and capital expenditure, but the relationship between those figures may matter more than any individual number.

Strong cloud growth accompanied by improving cash generation would support the argument that AI investment is producing durable value. Strong growth accompanied by another material increase in capital expenditure and weaker margins would be less reassuring.

The genuinely negative combination would be slower growth alongside continued increases in spending.

Cash flow is coming under pressure

Alphabet has already demonstrated the emerging problem.

Google Cloud revenue grew 82% in its second quarter, operating profit more than tripled and its backlog reached $514 billion. Demand was clearly not the issue.

However, Alphabet spent $44.9 billion on capital expenditure during the quarter and reported negative free cash flow of $5.9 billion. It also increased its 2026 capital-expenditure guidance to between $195 billion and $205 billion, with spending expected to rise significantly again in 2027.

Alphabet’s shares fell following the results despite the exceptionally strong operational performance. Investors were not questioning whether AI demand existed. They were questioning how much future cash flow would have to be reinvested merely to remain competitive.

Reuters estimates suggest Microsoft, Alphabet, Amazon, Meta and Oracle could collectively spend more on capital expenditure than they generate in free cash flow by 2027. Consensus capital-expenditure forecasts for the group have risen from approximately $485 billion at the start of 2026 to around $730 billion.

The AI giants are gradually becoming less like asset-light software businesses and more like capital-intensive infrastructure companies.

The build-out increasingly requires external finance

Oracle provides an even clearer example.

Its cloud-infrastructure revenue increased by 77% in its latest financial year, but free cash flow was negative $23.7 billion. Oracle raised $43 billion of debt and $5 billion of equity during the year and expects to raise approximately another $40 billion through debt and equity in its 2027 financial year.

This does not mean Oracle’s strategy will fail. Its contracted backlog is substantial, and revenue growth is accelerating. It does, however, illustrate the financing burden required to turn long-dated AI commitments into operational infrastructure.

AI-related borrowing has approached 15% of US investment-grade corporate issuance during 2026. Technology companies are increasingly using bonds, leasing structures and project finance to spread the cost of their data-centre investments.

The bond market is therefore becoming an increasingly important judge of the AI trade. Equity investors may focus on long-term addressable markets, but credit investors must assess cash flow, collateral, refinancing risks and whether customers can honour multi-year commitments.

Financing is becoming more circular

Another warning is the growing financial interdependence between AI suppliers and their customers.

Nvidia has reportedly discussed providing substantial financing guarantees for an OpenAI-related data-centre project, while also considering financing associated purchases of its own chips.

The arrangements may never proceed in their reported form, but the proposed structure is revealing. Nvidia could potentially use the financial strength generated from selling AI chips to help finance infrastructure that would purchase further Nvidia chips.

Vendor financing is not inherently problematic. It has supported the development of industries ranging from aviation to telecommunications.

However, it complicates the interpretation of demand. Investors must distinguish between independently financed customer demand and demand supported by suppliers, strategic investors or guarantees elsewhere within the same ecosystem.

The more circular the financing becomes, the more vulnerable the system may be to a reduction in confidence or the availability of capital.

Adoption is broad, but monetisation remains uneven

There is also a gap between AI experimentation and deep commercial implementation.

Recent research into S&P 500 companies estimated that only 11% had deeply integrated AI into their business processes by 2025, with a further 10% using it directly in producing goods or delivering services. Adoption had increased sharply, but the researchers found no measurable productivity difference between adopting and non-adopting companies at that stage.

That does not prove AI will fail to create productivity gains. Transformative technologies frequently require years of organisational change before the benefits become visible.

It does, however, expose a timing mismatch. Infrastructure expenditure is occurring immediately, while widespread productivity gains and customer returns may emerge much more gradually.

The trade is becoming more fragile

There are also signs that investor positioning has become stretched.

A July Bank of America fund-manager survey found that 82% of respondents regarded semiconductors as the market’s most crowded trade. Meanwhile, UBS expects hyperscaler capital-expenditure growth to slow from 76% in 2026 to 25% in 2027 and just 6% in 2028.

That would still represent enormous spending. However, current semiconductor valuations may require continued acceleration rather than simply a high absolute level of investment.

Pressure is appearing elsewhere in technology. Salesforce has weakened as investors question whether AI agents could disrupt traditional per-user software licensing before new AI revenues become sufficiently meaningful.

AI can therefore be transformative while simultaneously challenging the business models and profit pools of existing technology companies.

A warning light, not a dead canary

Bytes Technology is unlikely to be the single company that signals the end of the AI trade. Its challenges are partly company-specific, and Microsoft’s incentive changes could ultimately accelerate sales of Copilot and cloud services.

However, Bytes fits a broader pattern:

Revenue is rising, but margins are under pressure.

Demand is strong, but cash flow is weakening.

Infrastructure commitments are expanding, but so are borrowing requirements.

Adoption is increasing, but measurable customer returns remain uneven.

Suppliers are prospering, but some may increasingly need to help finance their own customers.

The next phase of the AI trade will therefore not be decided simply by whether artificial intelligence is transformative. It probably is.

It will be decided by whether AI revenues and productivity gains are sufficient to justify the capital expenditure, energy consumption, financing costs and depreciation required to deliver them.

What this means for MGTS Qualis Funds

This developing risk has directly influenced the positioning of MGTS Qualis Growth.

We have reduced the portfolio’s concentrated exposure to the Nasdaq and broadened the US allocation through both active smaller companies and global quality-dividend businesses. We have also reduced exposure to areas where recent returns have become increasingly dependent on a small number of AI, semiconductor and memory companies.

This does not represent a rejection of artificial intelligence. AI remains one of the most important long-term investment themes in global markets.

However, we do not believe investors should assume that every company participating in the AI ecosystem will earn an attractive return, or that today’s market leaders will capture the majority of tomorrow’s profits.

The role of diversification is particularly important when a dominant investment narrative becomes crowded, capital intensive and increasingly dependent on continued access to cheap financing.

Bytes may not be the dead canary in the AI coal mine. But it could be one of the first indications that the market is moving from rewarding AI expenditure at any price to demanding evidence of sustainable margins, cash flow and returns on capital.

This article is for information only and does not constitute investment advice or a personal recommendation. Capital is at risk, and the value of investments can fall as well as rise.