Technology companies are committing billions to AI infrastructure while restructuring their workforces. The connection is more complicated than machines replacing employees. It raises a larger question about how companies value computing capacity, human expertise and the capabilities they will need in the future.
TechiesJournal editorial illustration of the competing investment demands of AI computing infrastructure and human expertise. Illustrative, not a factual depiction of corporate decisions.
The investment shift we should be paying attention to
Oracle spent approximately $55.7 billion on capital expenditure in fiscal 2026, compared with $21.2 billion the previous year. Much of that increase supported data-centre expansion.
During the same period, its reported workforce declined from approximately 162,000 to 141,000 employees.
Microsoft has reportedly offered voluntary exits to around 60,000 eligible employees. Meta has reduced its workforce substantially while expanding AI infrastructure investment. Amazon, Block and other technology companies are also redesigning their organizations, although their reasons differ.
For years, we associated technology-company growth with expanding engineering teams, hiring more specialists and building larger organizations. Companies still need skilled people, but increasingly, growth also demands enormous investments in GPUs, data centres, networks, electricity and the infrastructure required to operate AI services.
I do not believe these developments can be explained simply by saying AI is replacing workers. That interpretation overlooks the economic decisions taking place behind the technology.
What interests me more is whether technology companies are changing how they decide where future business value will come from, and how much of that value they expect to create through computing infrastructure rather than expanding their workforce.
There are good reasons to make that shift. There are also consequences that may not appear immediately in financial results.
When software growth requires more physical infrastructure
Software companies have always invested in infrastructure. Cloud computing, enterprise platforms and online services depend on servers, storage, networking and data centres.
But developing and running advanced AI systems introduces a different scale of investment.
Training large models requires substantial computing resources. Serving those models to millions of customers creates continuing demand for accelerators, memory, networking and electricity. Some AI workloads consume significant computing capacity with every request, although costs vary by model, architecture and usage.
The financial question is whether the revenue and business value generated by those workloads justify building and operating the required infrastructure.
Oracle’s fiscal 2026 results illustrate the pressure. Its operating cash flow was approximately $32 billion, while capital expenditure reached $55.7 billion. The company spent more on capital investment than it generated from operations during the year.
That does not automatically indicate financial weakness. Oracle raised substantial financing, and infrastructure built today may generate revenue over many years. But the scale of investment means management must make careful decisions about where money goes.
For technology companies, workforce spending is part of that calculation.
Yet a company can change its workforce for different reasons. It may need to support infrastructure investment, use automation to reduce repetitive tasks or reorganize teams that have become unnecessarily complicated.
Understanding which of these forces is at work matters more than treating every workforce announcement as evidence of the same trend.
Different companies, different decisions
Meta provides one of the clearest examples of the connection between computing investment and workforce spending.
On Meta’s first-quarter call on 29 April 2026, Chief Financial Officer Susan Li linked the company’s growing infrastructure spending to its plan to reduce its employee base in May, saying a leaner operating model would help offset the substantial investments the company was making. The same call transcript records her describing how Meta weighs requests for headcount and for computing capacity.
Meta subsequently reported significant workforce changes. Its June 2026 results recorded 75,472 employees, including approximately 8,000 affected by a May reduction. By September 30, reported headcount was 66,775, down 15% from a year earlier.
These are actual reported employee counts, although the difference between two reporting dates should not be mistaken for an exact number of layoffs.
Microsoft illustrates a different situation. Its October 2026 programme made approximately 60,000 employees reportedly eligible for voluntary exits across participating organizations. The programme represents an option for employees to leave, not confirmation that all eligible employees will depart.
Salesforce offers another example. In a late-August 2025 podcast interview, its CEO Marc Benioff said he had reduced his customer-support headcount from about 9,000 to about 5,000 as AI agents took on more of the work. A Salesforce spokesperson told Fortune the company no longer needed to backfill support engineer roles and had redeployed hundreds of employees. These are leadership-reported figures for a particular function, not a measure of company-wide employment change, and the coverage reviewed gives no breakdown of how the roughly 4,000 positions were reduced.
Amazon’s January 2026 announcement followed yet another explanation: reducing management layers, simplifying decision-making and increasing accountability across teams.
Looking across these cases, the industry is not following a single formula.
Meta’s CFO has discussed an infrastructure-versus-workforce spending trade-off. Salesforce illustrates the effect of automation on a particular type of work. Amazon emphasizes organizational efficiency. Oracle combines major infrastructure expansion with restructuring that explicitly includes AI adoption.
These differences are essential to understanding the larger development.
Technology companies have reorganized for decades, often because of competitive pressure, changing products, excess hiring or inefficient structures. AI adds new capabilities and investment demands to those established business decisions.
The question is what kind of organization emerges from them.
Productivity is not the whole measure of engineering capability
Consider an engineering team using AI coding tools to complete development tasks faster.
Management may observe that fewer hours are needed to produce the same amount of code or complete a similar set of assignments. That can represent a genuine improvement.
But how should the organization use it?
There are at least three possible outcomes.
The first is that unnecessary work disappears. If automation removes repetitive tasks without damaging service quality, the business becomes more efficient.
The second is that the required skills change. Engineers may spend less time on routine implementation and more time evaluating designs, integrating systems, investigating behaviour and ensuring reliability. The work remains, but its composition changes.
The third possibility is more concerning: the organization reduces a capability it still needs because the value of that capability was not adequately measured.
The third outcome is the risk that interests me. It should not be confused with the first two.
Writing code is only part of engineering. An experienced professional also understands why a system behaves unexpectedly, how components interact and what happens when an apparently minor change creates a production failure.
Those contributions do not always appear in productivity measurements.
In AI Writes the Code. Who Understands What Happens in Production?, I examined why faster code generation does not remove the responsibility for operating and maintaining the resulting system.
The same reasoning applies to workforce decisions.
A smaller engineering team may perform better when it has clearer responsibilities, stronger expertise and effective tools. But if management removes people whose knowledge is still essential, the organization may discover the cost only when an unfamiliar problem arises.
The employee count cannot tell us which outcome has occurred.
Restructuring can also make an organization stronger
There is a serious argument in favour of the changes many technology companies are making.
Large organizations often accumulate unnecessary processes and management layers. Teams duplicate responsibilities, decisions move through too many approvals, and employees spend time coordinating work that could be organized more effectively.
Eliminating those inefficiencies can improve productivity and make employees more effective.
AI creates additional opportunities to automate routine activities and direct human effort toward problems requiring greater judgment. Infrastructure expansion also generates demand for expertise in hardware, networking, energy, security and data-centre operations.
Preserving an existing organizational structure simply because it once supported growth is not necessarily responsible management.
In fact, refusing to adapt can damage an organization just as much as changing it carelessly.
The stronger question is whether the new operating model delivers better results over time.
If a company can serve customers more reliably, respond to problems faster and develop the expertise it needs with fewer people, that is a meaningful improvement.
But efficiency should be demonstrated through the performance of the organization, not merely the size of its payroll.
Human expertise has a different investment cycle
When a company builds a data centre, it can calculate equipment costs, capacity, expected utilization, depreciation and potential returns.
Developing human expertise is much harder to measure in the same way.
An engineer becomes capable of handling difficult situations through experience: projects that fail, incidents that require investigation, technical decisions that prove imperfect and guidance from more experienced colleagues.
Over time, that experience becomes part of the organization’s ability to solve problems.
Oracle’s fiscal 2026 annual report warns that restructuring can cause a loss of valuable institutional knowledge, shortages of sufficiently skilled employees in certain roles, and increased restructuring costs and reduced productivity.
That disclosure matters. It shows that the potential loss of organizational expertise is a recognized business risk, even within companies actively restructuring.
It does not prove that Oracle has suffered every one of those consequences. Risk disclosures describe potential outcomes, not necessarily observed failures.
There is also a longer-term question.
In When AI Does the Junior Work, Who Develops the Next Generation of Professionals?, I examined how automating routine assignments can affect the practical experience through which less-experienced professionals develop judgment.
Now consider that issue alongside workforce restructuring.
Companies may expect smaller teams to manage increasingly complex AI-enabled systems. Those teams will need people who understand architecture, security, reliability and the behaviour of connected systems.
Where will that expertise come from?
AI-assisted learning may create better ways to develop skills. New technical roles will emerge, and some existing career paths will change considerably.
But experience still has to be developed through meaningful responsibility.
In Human Sustainability in the AI Age, I explored the importance of preserving human capability and opportunity as machines become more capable.
For employers, that principle has a practical dimension. The future workforce cannot consist only of experienced professionals recruited from somewhere else. Organizations must help develop the people they will eventually need.
The full return on AI investment
The economics of AI should include more than the cost of a model and the labour it might replace.
An automated business process still requires infrastructure, integration, security, monitoring, testing and operational support. Depending on the application, those responsibilities may become more demanding as AI is introduced.
When something goes wrong, the organization must still be able to identify the problem, understand its effects and recover safely.
In Navigating the AI World, I discussed why AI capabilities depend on the systems and engineering practices surrounding the model.
That broader operating environment should also be considered when calculating productivity.
A company may save money by automating certain activities while needing to invest more in other kinds of expertise. That is not a failure of automation. It is part of designing an effective operating model.
The quality of an AI investment should ultimately be judged by the business capability it creates: what the organization can deliver, how reliably it can operate and whether the economics remain sustainable.
Workforce decisions are inseparable from that outcome, even when they cannot be reduced to a simple formula.
The investment decision that will matter later
The technology industry has reasons to be optimistic about AI. Computing infrastructure is expanding, AI capabilities are improving and businesses are discovering new ways to use them.
Some existing roles will contract. Others will become more important, and entirely new forms of work will emerge.
I do not believe technology companies should preserve every position or hesitate to invest in computing capacity simply because organizational change is difficult.
But physical computing capacity and human capability are both long-term requirements, and companies may evaluate them using very different measures and time horizons.
A data centre requires years of financial planning. An experienced engineering organization also takes years to develop, even though much of its value is less visible in an investment spreadsheet.
The strongest technology companies will not necessarily be those with the largest workforces or the greatest amount of AI infrastructure. They will be those that understand what each investment enables, what expertise their systems require and how to sustain that expertise as technology changes.
That is the question I would like technology leaders to consider as they plan the next wave of AI investment.
We are spending enormous amounts to build the computing capacity of the future. Are we giving equal thought to developing the human capability needed to make that future work?
How major technology companies are reallocating investment and work
The comparison below brings together different examples of infrastructure investment, AI-enabled productivity and organizational restructuring. It supports the discussion above but should not be read as a ranking of AI-related layoffs.
All monetary values are in US dollars. Reported headcount, announced reductions, voluntary exit eligibility and spending forecasts are different measures and must not be added together.
What kind of number is each one?
The same eight companies as the table below, regrouped by the kind of number their sources give. Each cell repeats a figure from the table.
| Company | Headcount in filings or results | Reduction the company stated | Offer to leave | One function, stated by an executive | Spending figure |
|---|---|---|---|---|---|
| Microsoft | about 223,000, Jun 30, 2026 | not given in the table | about 60,000 eligible, press-reported (not departures) | not given in the table | not given in the table |
| Oracle | about 162,000 to 141,000, May 2025 to May 2026 | not given in the table | not given in the table | not given in the table | capex $21.2B to $55.7B, fiscal 2025 to 2026 |
| Meta | 75,472 to 66,775, Jun 30 to Sep 30, 2026 | about 8,000 affected by the May 2026 reduction | not given in the table | not given in the table | not given in the table |
| Amazon | not given in the table | about 16,000 corporate roles, Jan 28, 2026 | not given in the table | not given in the table | not given in the table |
| Block | not given in the table | about 4,000 positions, Feb 26, 2026 | not given in the table | not given in the table | not given in the table |
| Salesforce | not given in the table | not given in the table | not given in the table | support staff about 9,000 to 5,000, Aug 2025 | not given in the table |
| Cisco | not given in the table | fewer than 4,000 jobs, under 5%, May 2026 | not given in the table | not given in the table | not given in the table |
| Alphabet | not given in the table | not given in the table | not given in the table | not given in the table | 2026 capex forecast $195B to $205B |
Read across a row to see what one company gave. Read down a column to see who gave the same kind of number. Even within a column the figures differ in date, scope and base, which is why they are not added together.
Read one layer at a time
The table below separates what each company reported, what it or its reporters explained, and the author’s interpretation. Pick a layer to compare it across all eight companies. Nothing is removed from the page.
| Company | Reported figures and period | Documented explanation | Author’s interpretation |
|---|---|---|---|
| Microsoft | Approximately 60,000 employees reportedly eligible for voluntary exits in October 2026 (press-reported, confirmed to reporters by Microsoft). Approximately 223,000 employees as of June 30, 2026 (Form 10-K). | Microsoft’s October 26 announcement sets the rules (level 70 and below, participating teams in the US, Canada, the UK and India) but gives no headcount. TechCrunch reports that Microsoft confirmed about 60,000 eligible. The 223,000 figure is from Microsoft’s fiscal 2026 Form 10-K. | A major workforce-planning change during AI expansion, but eligibility is not completed job loss, and the financial relationship to infrastructure spending is not quantified. |
| Oracle | Workforce approximately 162,000 to 141,000 between May 2025 and May 2026. Annual capital expenditure $21.2B to $55.7B. | FY2026 Form 10-K reports the 141,000 employees, the capital expenditure, data-centre expansion and restructuring measures that include AI adoption. The 162,000 figure is from the FY2025 Form 10-K. | A strong case of simultaneous capital expansion and workforce contraction. The net employee change cannot be attributed entirely to AI or infrastructure financing. |
| Meta | 75,472 employees at June 30, 2026, including approximately 8,000 affected by May reductions. 66,775 employees at September 30, down 15% year over year. | Second-quarter results and third-quarter results report headcount. Meta’s CFO separately linked infrastructure spending to the planned reduction on the first-quarter earnings call. | An unusually direct example of an investment trade-off discussed by the company’s finance chief, with subsequent disclosed workforce changes. Headcount differences are not equivalent to layoff counts. |
| Amazon | Approximately 16,000 corporate roles cut, announced January 28, 2026. | Amazon’s official statement cites fewer management layers, greater ownership and reduced bureaucracy, alongside continued strategic hiring. | An organizational-efficiency case that cautions against attributing every reduction to direct AI replacement. |
| Block | Announced cuts of about 4,000 positions on February 26, 2026, taking headcount from over 10,000 to just under 6,000, roughly 40%. | Fortune reports Jack Dorsey crediting intelligence tools and smaller, flatter teams, and notes that he also tied the cuts partly to pandemic-era hiring. Reuters also reported the announcement. | A direct example of management expecting AI tools to support a materially smaller organization, rather than mainly financing data-centre construction, although part of the stated reasoning concerned earlier hiring. |
| Salesforce | Customer-support headcount reportedly reduced from approximately 9,000 to 5,000, as described by its CEO in late August 2025. | Fortune and The Register report Benioff’s podcast remarks, and a spokesperson’s statement that the company stopped backfilling support engineer roles and redeployed hundreds of employees. | Shows AI affecting staffing within a particular business function. The figures come from the CEO’s remarks rather than an audited role-level headcount series. |
| Cisco | Announced a reduction of fewer than 4,000 jobs, or less than 5% of employees, in May 2026. | Cisco’s official statement connects organizational change with strategic investment priorities and AI-related demand. | Illustrates reallocation toward different technical capabilities, including within a company supplying the infrastructure needed for AI expansion. |
| Alphabet (Google) | Full-year 2026 capital expenditure forecast of $195B to $205B, raised in July 2026 and left unchanged at the Q3 report on October 29, 2026. | Alphabet’s Q3 earnings discussion confirms that the guidance remained unchanged as of October 29. | Demonstrates the financial scale of AI infrastructure investment, without establishing a corresponding reduction in workforce. |
References and further reading
Primary company disclosures
- Oracle Corporation, Fiscal 2025 Form 10-K, fiscal year ended May 31, 2025. SEC filing. Workforce at May 31, 2025.
- Oracle Corporation, Fiscal 2026 Form 10-K, fiscal year ended May 31, 2026. SEC filing. Workforce, capital expenditure, restructuring and institutional-knowledge risks.
- Meta Platforms, First Quarter 2026 Earnings Call Transcript, April 29, 2026. Transcript (PDF). The CFO’s remarks linking infrastructure spending to the planned May reduction.
- Meta Platforms, Second Quarter 2026 Results, July 2026. Investor release. Workforce reporting and the May restructuring.
- Meta Platforms, Third Quarter 2026 Results, October 21, 2026. Investor release. Updated workforce and AI infrastructure outlook.
- Microsoft, Fiscal 2026 Form 10-K, fiscal year ended June 30, 2026. SEC filing. Total employees as of June 30, 2026.
- Microsoft, Introducing a Voluntary Exit Offer, October 26, 2026. Official announcement. Programme eligibility and stated workforce rationale.
- Amazon, Update on Our Organization, January 2026. Official announcement. Organizational restructuring and affected roles.
- Cisco, Our Path Forward, May 13, 2026. Official statement. Workforce realignment and strategic priorities.
Independent reporting and supporting evidence
- TechCrunch, Microsoft Offers Buyouts to Around 60,000 Employees, October 26, 2026. Article. Reporting on programme scale.
- Fortune, Block, Jack Dorsey and the 4,000 job cuts, February 27, 2026. Article. Leadership’s AI-productivity rationale and the pandemic-hiring context.
- Reuters, Block Announces Major Workforce Reductions, February 26, 2026. Article. Report of the announcement.
- Fortune, Salesforce CEO on AI agents and support staffing, September 2, 2025. Article. Benioff’s remarks and the company spokesperson’s statement.
- The Register, Salesforce support headcount, September 2, 2025. Article. Second report of the same remarks.
- Alphabet, Q3 2026 Earnings Discussion, October 29, 2026. Transcript hosted by The Motley Fool. Reaffirmed 2026 capital expenditure guidance.
Related TechiesJournal publications
- Human Sustainability in the AI Age, Human capability and opportunity in an AI-driven economy.
- Navigating the AI World, AI infrastructure and operational responsibilities.
- When AI Does the Junior Work, Who Develops the Next Generation of Professionals?, Automation and professional development.
- AI Writes the Code. Who Understands What Happens in Production?, Engineering responsibility beyond AI-generated code.
Evidence reviewed through October 30, 2026. Company reporting periods, financial definitions and workforce measures differ. The article distinguishes documented facts, management explanations and the author’s interpretation. It does not establish that AI infrastructure investment directly caused every reported workforce reduction.