The AI Spending Boom Does Not Prove Your Tax Career Is Obsolete

Last reviewed: July 26, 2026.

The original version of this article claimed that the AI industry had borrowed $3 trillion, owed about $100 billion in annual interest, and therefore needed to replace 10 million white-collar workers every year.

That chain cannot be supported.

AI spending includes equity, operating expense, capital expenditure, debt, data-center construction, chips, energy, and investments made by companies that also have large non-AI businesses. There is no credible accounting rule that converts the total into a required number of displaced workers.

The AI investment boom is real. The “10 million replacements” arithmetic was not.

What verified investment data can tell us

Public filings show that major technology companies are spending heavily on technical infrastructure. Alphabet reported that capital expenditures rose from $52.5 billion in 2024 to $91.4 billion in 2025 and said it expected substantially higher technical-infrastructure investment in 2026.

That spending can support many business models: cloud services, consumer products, advertising, enterprise software, scientific computing, and internal productivity. It is not a pure bet on labor replacement.

Capital expenditure also does not equal debt. A company can finance investment through operating cash flow, equity, debt, or a mixture. Any article that treats all AI investment as borrowed money should show the balance-sheet evidence company by company.

What current adoption data says

The Census Bureau’s nationally representative Business Trends and Outlook Survey found that overall business AI use hovered around 17% to 20% from December 2025 through May 2026.

In the detailed 2026 AI supplement:

• 18% of firms reported using AI in at least one business function;
• the employment-weighted share was 32%;
• 57% of adopting firms used AI in three or fewer business functions; and
• common uses included sales and marketing, strategy, and information technology.

That is fast diffusion. It is not universal deployment, and adoption is not the same thing as headcount reduction.

Employment effects remain uncertain

BLS explicitly says employment trajectories for many AI-susceptible occupations remain uncertain. Some occupations exposed to generative AI are still projected to grow.

For the tax-related occupations most relevant to EA candidates, current 2024–2034 projections show:

• accountants and auditors: 5% growth, with about 124,200 openings per year;
• tax preparers: 4.5% growth, with about 10,400 openings per year.

Projections can change, and they do not guarantee an individual career outcome. They do contradict the claim that official labor forecasts already show tax work collapsing.

AI will change tax work

Rejecting bad replacement arithmetic is not the same as denying disruption.

AI can accelerate:

• document intake and classification;
• first-draft client communications;
• research retrieval;
• reconciliation and anomaly detection;
• workpaper preparation; and
• routine review.

Those changes can reduce time spent on some tasks and reshape entry-level training. Firms may need fewer hours for a particular workflow even if total demand grows.

The right question is not “Will AI touch tax work?” It already does. The better question is “Which responsibilities remain with an accountable person, and what skills become more valuable when routine work gets faster?”

Tax preparation has an accountability layer

IRS rules require compensated federal return preparers to have individual PTINs. The signing preparer is the individual with primary responsibility for overall preparation accuracy.

Paid preparers may face penalties for unreasonable positions, willful or reckless conduct, failing to sign, or failing to furnish a PTIN. EAs also remain subject to Circular 230 standards when practicing before the IRS.

Software can contribute to a work product. It does not become the enrolled practitioner, accept professional discipline, or replace the taxpayer’s and preparer’s legal responsibilities.

That accountability layer does not freeze workflows in place. It means a firm adopting AI still needs people who can verify facts, identify missing information, interpret authority, explain uncertainty, and own the final professional decision.

The career takeaway

Do not build a career plan on either extreme:

• “AI spending guarantees millions of immediate job losses.”
• “Professional credentials make automation irrelevant.”

The evidence supports a middle position. AI adoption is material and uneven. Work will be reorganized. Some tasks will shrink. New tools will raise expectations. Meanwhile, current BLS projections still show tax and accounting openings, and IRS rules continue to assign preparation and representation responsibilities to people.

For an EA candidate, the practical response is to become harder to commoditize:

• know the underlying tax rule;
• verify model output against primary authority;
• get good at incomplete-fact interviews;
• document judgment;
• learn representation procedure; and
• use automation without outsourcing accountability.

Huge investment can accelerate change. It cannot, by itself, prove your profession’s expiration date.

Practice the federal tax rules AI output still needs checked against →

---

Sources: Alphabet filing discussing 2025–2026 technical-infrastructure investment · Census Bureau business AI-use analysis · Census AI diffusion working paper · BLS AI impacts in employment projections · BLS occupational projections · IRS PTIN FAQ

Related: AI Isn't Killing Tax Preparation · Chinese Open-Weight AI Weakens the Monopoly Story · Is the EA Credential Worth It?