IBM 10-K Analysis: Why Real Earnings Growth Is 8%, Not 76%

IBM, the $67.5 billion hybrid cloud and AI company, reported a 76% net income surge in FY2025 — its best earnings growth in over a decade. Two weeks later, the stock crashed 13%, erasing $31 billion in market cap, when Anthropic announced Claude Code could automate the COBOL modernization work that underpins IBM's $21 billion consulting segment. The 10-K reveals both reactions were built on numbers that don't hold up.

The headlines wrote themselves: 76% earnings growth, a revitalized IBM, an AI transformation story finally bearing fruit. Analysts debated whether IBM had reached an inflection point. Then the Claude Code panic hit, and the same analysts scrambled to quantify consulting's AI exposure. Neither side looked at what the filing actually said about where the 76% came from — or how much of IBM's profit consulting actually represents.

The 10-K tells a different story. Two non-recurring events — a $2.0 billion pension settlement that crushed FY2024 results and a $1.435 billion Q4 tax windfall that inflated FY2025 — created a $3.1 billion artificial tailwind. Strip both out, and IBM's true earnings growth was approximately 8-9%. Meanwhile, consulting is only 17.1% of IBM's segment profit. The real transformation happening inside IBM is a quiet software takeover that generates 61.7% of total segment profit at 31.1% margins — a story that neither the earnings euphoria nor the Claude Code panic captured.

The 76% Phantom: Earnings Quality Forensics

IBM's FY2025 net income of $10.593 billion looks transformative on the surface — a 75.9% surge over FY2024's $6.023 billion. But the comparison is meaningless for valuation because both years are distorted in opposite directions.

In Q3 2024, IBM transferred a portion of its pension obligations to an insurer, booking a one-time charge that depressed the prior-year baseline:

"The transfer to an insurer of a portion of the Qualified PPP's defined benefit pension obligations and related plan assets which resulted in a pension settlement charge of $2.7 billion ($2.0 billion net of tax) in the third quarter of 2024."

That $2.0 billion net-of-tax charge made FY2024 look artificially weak — the base from which FY2025's "76% growth" is measured. Then in Q4 2025, the opposite happened: IBM resolved $2.67 billion in unrecognized tax benefits, producing a $1.435 billion tax windfall that inflated the current year. The filing's own tax disclosures reveal the scale:

"The amount of unrecognized tax benefits at September 30, 2025 is $9,325 million which can be reduced by $681 million associated with timing adjustments, potential transfer pricing adjustments, and state income taxes. The net amount of $8,644 million, if recognized, would favorably affect the company's effective tax rate."

By December 31, that $9,325 million had dropped to $6,655 million — a $2,670 million decrease in a single quarter. The Q4 effective tax rate was approximately -34.6%, compared to the 9-month operational rate of 19.3%. This is the definition of non-recurring.

The investment implication is straightforward: investors must re-anchor valuation on approximately $8.7 billion in normalized net income (26.5× P/E at ~$246), not the $10.6 billion GAAP figure (21.7× P/E). More critically, FY2026 reported earnings will likely decline 15-20% year-over-year as the effective tax rate reverts to 16-19%, even if underlying operations continue to improve. At a normalized 19% tax rate on roughly $10.5 billion in pretax income (assuming 5% operational growth), FY2026 net income would be approximately $8.5 billion versus $10.6 billion reported in FY2025 — an apparent 20% decline that obscures real improvement.

The cash flow statement confirms the distortion. Operating cash flow of $13.2 billion actually declined 1.9% year-over-year while net income surged 76%, and free cash flow per share fell 3.6% — a 77 percentage-point divergence from EPS growth. When the income statement says one thing and the cash flow statement says the opposite, follow the cash. IBM's FY2025 net income surge of 76% was inflated by $3.1 billion in non-recurring tailwinds — a 2024 pension settlement and a 2025 tax windfall — reducing true underlying earnings growth to approximately 8-9%.

Software's Quiet Takeover: The Real Bull Thesis

If the 76% headline is a phantom, the real question is what IBM is worth on normalized earnings. The answer depends on a structural shift that most coverage overlooks: IBM is already a software-first company by profit.

In the first nine months of 2025, IBM's software segment generated $6,517 million in segment profit on $20,932 million in revenue — a 31.1% margin. This single segment accounted for 61.7% of IBM's total segment profit, producing $0.311 in profit for every revenue dollar. By comparison, consulting generated $0.115 per dollar — making software 2.7 times more profitable on a per-dollar basis.

This profit concentration is driving measurable operating leverage. Operating income grew 29.1% on just 7.6% revenue growth — a 3.8× leverage ratio that shows revenue is flowing to higher-margin products. Software revenue itself grew 9.2% for the nine months (8.4% at constant currency), accelerating to 14% in Q4 2025, the fastest annual growth rate in IBM's modern history. Gross margins in software reached 83.5%, providing significant room for incremental profit as the segment scales through Red Hat, watsonx, and the newly integrated HashiCorp portfolio.

The complication is that approximately $1.8 billion per year in acquisition-related charges (amortization of acquired intangibles from HashiCorp and prior deals) sits between GAAP and non-GAAP operating income — a 17% gap. These charges are the direct, recurring cost of IBM's M&A-driven software strategy. They obscure the software segment's true contribution to the bottom line but will persist as long as IBM continues acquiring. The Confluent acquisition adds another $11 billion in purchase price with a likely similar goodwill allocation. Investors using IBM's non-GAAP numbers are implicitly assuming these costs are temporary; they are not.

The bull thesis for IBM is not the AI narrative or the 76% headline — it is the software mix shift. IBM's software segment generates 61.7% of total segment profit at 31.1% margins, producing 2.7 times more profit per revenue dollar than consulting and delivering 3.8× operating leverage in FY2025. If software continues to grow at high single digits while consulting holds steady, IBM's normalized earnings trajectory could reach the 10-12% range — not spectacular, but enough to support a 22-24× multiple on normalized earnings.

The Consulting Mirage: Why the $31 Billion Panic Was Wrong

When Anthropic announced that Claude Code could automate COBOL modernization, IBM's stock dropped 13% in a single session, erasing $31 billion in market value. The market was pricing in the destruction of IBM's consulting business — and it was wrong by a factor of roughly five.

Consulting contributed only 17.1% of IBM's total segment profit in the first nine months of 2025 at an 11.5% margin. Even in a catastrophic scenario where consulting revenue declines 20%, total segment profit would fall by approximately 3.4%. The $31 billion market cap loss — representing roughly 13% of IBM's total value — priced in destruction far exceeding consulting's actual economic contribution.

But dismissing the AI threat entirely would be premature. The filing reveals a more nuanced signal that the market reaction missed:

"For the three and nine months ended September 30, 2025, Consulting signings decreased 4.1 percent as reported (5.1 percent adjusted for currency) and 10.1 percent as reported (11.0 percent adjusted for currency), respectively. However, the quality of our signings continued to strengthen, with more strategic signings with new clients and expanded engagements with existing clients."

Signings are a leading indicator of revenue, typically by 6-12 months. The 10.1% decline in 9M 2025 signings — despite management's insistence that "quality" is improving — suggests consulting revenue growth could turn negative in the second half of 2026. The trailing book-to-bill ratio remains above 1.0×, and consulting segment margins actually improved from 9.9% in FY2024 to approximately 12.2% through 9M 2025, a 220 basis-point expansion indicating IBM is shedding low-margin work for higher-value engagements. This is a volume story, not a margin story — the distinction matters for assessing long-term damage.

The $64 billion remaining performance obligation backlog, with 68% ($43.5 billion) expected to be recognized in the next two years, provides significant runway before any AI-driven disruption could create an actual revenue cliff. Most consulting and software contracts are multi-year with termination protections. IBM's consulting segment contributes only 17.1% of total segment profit, meaning even a 20% revenue decline from AI disruption would reduce total profit by just 3.4% — far less than the $31 billion market reaction implied. The Claude Code panic created a potential mispricing, but signings trajectory — not the AI headline — is the canary in the coal mine for FY2026 consulting revenue.

The $70 Billion Leverage Bet

IBM's most significant strategic shift is not the AI pivot — it is the capital allocation reversal. After accumulating $170.6 billion in historical share buybacks, IBM returned zero capital via repurchases in FY2025, allowing shares outstanding to increase by 1.2%. Instead, the company deployed $7.4 billion on HashiCorp and committed $11 billion for Confluent — the largest acquisition spree in IBM's modern history.

The HashiCorp deal reveals the economic reality of this pivot. The actual purchase price was $7,433 million — $1 billion more than the commonly cited $6.4 billion. Of this total, $4,683 million (63%) was allocated to goodwill, the highest proportion among IBM's recent acquisitions. The remaining identifiable intangible assets ($1,835 million in client relationships, technology, and trademarks) carry 5-13 year amortization, creating approximately $230-310 million per year in acquisition-related charges that flow through the income statement.

This M&A-driven growth compounds IBM's already elevated balance sheet risk. Total debt stood at $61.3 billion at year-end 2025, with $5.8 billion in U.S. dollar debt maturing in 2026 at a 3.7% weighted average rate and $4.1 billion maturing in 2027 at 3.3%. Refinancing this $9.9 billion at current market rates of 5% or higher would increase annual interest expense by approximately $150-250 million. The Confluent acquisition, which closed on March 17, 2026, could push total debt toward $70 billion or higher.

The goodwill comparison tells the story. IBM's $67.7 billion in goodwill represents 44.6% of total assets — 2.2 to 3.3 times the peer average. A hypothetical 20% impairment would produce a $13.5 billion charge, exceeding FY2025 normalized net income. As IBM continues to shift from buybacks to acquisitions, this concentration will only increase: HashiCorp alone added $4.7 billion in goodwill, and Confluent will add a comparable amount.

The acquisition strategy is generating returns so far — return on incremental invested capital (ROIIC) of 50.4% and ROIC of 12.6% both exceed the 3.3% cost of debt. But interest coverage of 5.6× is the lowest among peers (Mastercard: 26.2×, Visa: 40.7×, ServiceNow: debt-free), leaving minimal room for margin compression or integration delays. IBM's capital returned consisted solely of $6.255 billion in dividends, with the 5-year dividend growth CAGR at a negligible 0.6% — effectively negative in real terms.

What the Price Assumes

At approximately $246 per share, IBM trades at 21.7× GAAP earnings — a number that flatters the story because it includes the $1.435 billion Q4 tax windfall. On normalized earnings of $8.7 billion, the P/E expands to 26.5×. For a company growing earnings at ~8-9% organically, this multiple requires an acceleration thesis.

A 26.5× multiple on normalized earnings requires 12-15% annual earnings growth over the next five years to justify a premium to the market. The filing supports approximately 8-9% organic growth, leaving a 4-6 percentage-point gap that must come from HashiCorp and Confluent synergies. FY2026 is the proof year — and it will be a triple-headwind year: the effective tax rate reverts to 16-19% (making reported earnings appear to decline 15-20%), Confluent integration costs hit the P&L, and $5.8 billion in debt requires refinancing at higher rates.

The free cash flow picture adds another layer of complexity. IBM's self-reported FCF of $14.7 billion includes adjustments for financing receivables that standard FCF calculations do not:

"A key objective of the Financing business is to generate strong returns on equity, and our Financing receivables are the basis for that growth. Accordingly, management considers Financing receivables as a profit-generating investment, not as working capital."

On a standard GAAP basis (operating cash flow minus capital expenditures), free cash flow was $12.1 billion — a $2.6 billion gap from IBM's self-reported figure. IBM's FY2026 FCF guidance of $15.7 billion is built on the company's broader definition and implicitly assumes the $1.4 billion tax benefit tail wind does not need to repeat. The organic FCF run rate is closer to $13.3 billion, meaning IBM needs $2.4 billion in incremental FCF to hit guidance — achievable but requiring flawless execution on both acquisitions, continued software growth, and z17 mainframe cycle tailwinds.

What to Watch

Five metrics will determine whether IBM's acquisition-driven thesis delivers or breaks:

1. Effective tax rate (Q1 FY2026): If the ETR stays below 10%, the tax benefit may be more structural than assumed and normalized earnings are higher. If it reverts to 16-22%, FY2025's headline was a phantom. Falsification point.
2. Consulting signings (Q4 FY2025 / Q1 FY2026): If the decline accelerates beyond -10%, consulting revenue turns negative in H2 2026. If signings turn positive, the Claude Code panic was a buying opportunity. Falsification point.
3. Software segment margin: Above 30% confirms operating leverage is intact. Below 28% for two consecutive quarters breaks the bull thesis. Confirmatory signal.
4. Total debt post-Confluent: Whether total debt crosses $70 billion and how IBM communicates the refinancing plan for the $9.9 billion maturity wall.
5. HashiCorp contribution: Management expects EBITDA accretion by year one (February 2026). If missed, goodwill impairment risk escalates.

At $246, the market implies IBM can deliver 12-15% annual earnings growth for five years. The filing supports the software transition as a genuine structural accelerant — 61.7% of profit at 31.1% margins with 3.8× operating leverage — but complicates the growth story with a $9.9 billion debt maturity wall, FY2026 tax normalization that will produce an apparent earnings decline, and $18.4 billion in acquisition commitments that leave zero margin of safety if either integration stumbles. IBM trades at 26.5 times normalized earnings of $8.7 billion, requiring 12-15% annual growth for five years, but the filing supports only 8-9% organic growth — leaving a gap that depends entirely on acquisition execution.

Frequently Asked Questions

What was IBM's real earnings growth in FY2025?

IBM's FY2025 GAAP net income of $10.593 billion represented a headline 75.9% increase over FY2024's $6.023 billion. However, this comparison is distorted by two non-recurring events: a $2.0 billion net-of-tax pension settlement charge that depressed FY2024 results, and a $1.435 billion Q4 2025 tax benefit from the resolution of $2.67 billion in unrecognized tax benefits. Adjusting both years, normalized FY2024 net income was approximately $8.0 billion and normalized FY2025 net income was approximately $8.7 billion — representing true underlying growth of approximately 8-9%.

Why was IBM's effective tax rate negative in FY2025?

IBM's FY2025 effective tax rate of -2.3% resulted from a Q4 2025 tax benefit of $1.435 billion, driven by the resolution of unrecognized tax benefits that decreased from $9,325 million at September 30 to $6,655 million at December 31, 2025. This $2,670 million decrease stemmed from the expiration of statutes of limitations and/or resolution of tax audit matters. IBM's underlying operational tax rate, based on the first nine months of 2025, was 19.3%.

How much of IBM's profit comes from consulting, and what is the real AI risk?

IBM's consulting segment contributed only 17.1% of total segment profit in the first nine months of 2025, generating $1,807 million at an 11.5% margin on $15,706 million in revenue. Software contributed 61.7% ($6,517M) at 31.1% margins. Even a catastrophic 20% decline in consulting revenue would reduce total segment profit by approximately 3.4%. However, consulting signings declined 10.1% in 9M 2025 — a leading indicator of potential revenue weakness 6-12 months forward.

What is IBM's software segment's actual profitability?

IBM's software segment generated $6,517 million in segment profit on $20,932 million in revenue during the first nine months of 2025, a 31.1% margin. Software's gross margin was 83.5%. For every dollar of revenue, software generates $0.311 in profit versus $0.115 for consulting — a 2.7× profit intensity ratio. Software revenue grew 9.2% for the nine months and 14% in Q4 2025, the fastest annual growth in IBM's history.

How does IBM's FCF differ from the standard calculation?

IBM defines free cash flow as operating cash flow minus the change in financing receivables minus net capital expenditures, plus software investment and asset sales. This produced $14.7 billion in FY2025. Standard FCF (operating cash flow minus CapEx) was $12.1 billion — a $2.6 billion gap. IBM treats its $11.4 billion financing receivables portfolio as "profit-generating investments, not working capital." Investors comparing IBM's FCF to peers should use the GAAP definition ($12.1B). At ~$246/share, EV/FCF is 23.0× GAAP versus 18.9× using IBM's definition.

What is IBM's debt situation after the Confluent acquisition?

IBM carried $61.3 billion in total debt at year-end 2025, with $5.8 billion in U.S. dollar debt maturing in 2026 at a 3.7% weighted average rate and $4.1 billion maturing in 2027 at 3.3%. Refinancing this $9.9 billion at current market rates (~5%+) would increase annual interest expense by $150-250 million. The Confluent acquisition ($11 billion, closed March 17, 2026) could push total debt toward $70 billion+. Interest coverage of 5.6× is the lowest among peers (MA: 26.2×, V: 40.7×).

How does IBM compare to ServiceNow, Mastercard, and Visa?

IBM trades at the lowest P/E (21.7× GAAP, 26.5× normalized) but also has the lowest revenue growth (7.6% vs NOW 20.9%, MA 16.4%, V 11.3%). IBM's FCF margin (17.9%) is roughly half ServiceNow's (34.5%) and one-third of the payment networks' (MA 52.3%, V 53.9%). IBM is the only peer with declining FCF per share (-3.6%), the highest goodwill concentration (44.6% vs 13-21% for peers), and the lowest interest coverage (5.6×). IBM's 2.3% dividend yield is unique among the peer set.

What impact does the HashiCorp acquisition have on IBM's financials?

IBM completed the HashiCorp acquisition in February 2025 for $7,433 million — not the commonly cited $6.4 billion. Of the purchase price, $4,683 million (63%) was allocated to goodwill. Identifiable intangible assets totaled $1,835 million with 5-13 year amortization creating approximately $230-310 million per year in amortization expense. Management expects HashiCorp to be accretive to Adjusted EBITDA within the first full year and accretive to FCF in year two.

Will IBM's FY2026 earnings grow or decline?

Based on filing evidence, IBM's reported FY2026 GAAP earnings are likely to decline 15-20% year-over-year, even if operations improve. The Q4 2025 tax windfall ($1.435B) will not recur, and the effective tax rate should revert to 16-19%. At a normalized 19% tax rate on approximately $10.5 billion pretax income (assuming 5% operational growth), FY2026 net income would be approximately $8.5 billion versus $10.6 billion reported in FY2025 — an apparent 20% decline that obscures underlying operational improvement.

What is IBM's RPO backlog and what does it signal?

IBM's remaining performance obligations totaled $64 billion at September 30, 2025, with 68% ($43.5 billion) expected to be recognized as revenue in the following two years. This represents approximately 95% of annual revenue and provides strong forward visibility. The RPO backlog counters the narrative that AI disruption could cause a sudden revenue cliff — most consulting and software contracts are multi-year with termination protections.

How has IBM's capital allocation changed?

IBM's FY2025 represents a fundamental break from historical capital allocation. The company returned zero capital via share buybacks after accumulating $170.6 billion in historical repurchases. Shares outstanding increased 1.2% (net dilution). Total capital returned consisted solely of $6.255 billion in dividends, with the 5-year DPS CAGR at just 0.6%. Meanwhile, IBM deployed $7.4 billion on HashiCorp and committed $11 billion for Confluent — the most significant strategic pivot from capital return to acquisitive growth in IBM's modern history.

What would prove this analysis wrong?

Three data points would invalidate the thesis: (1) If IBM's FY2026 effective tax rate remains below 10% in Q1-Q2, the tax benefit may be more structural than assumed, and normalized earnings are higher than $8.7B. (2) If software segment profit margin drops below 28% for two consecutive quarters, the operating leverage thesis breaks. (3) If consulting signings return to positive growth in Q4 2025/Q1 2026, the AI cannibalization narrative was premature and the market reaction was entirely macro-driven.

What is IBM's goodwill impairment risk?

IBM carries $67.7 billion in goodwill representing 44.6% of total assets — 2.2 to 3.3 times the peer average (NOW 13.7%, MA 17.6%, V 20.5%). A hypothetical 20% impairment would produce a $13.5 billion charge, exceeding FY2025 normalized net income of $8.7 billion. The HashiCorp acquisition added $4.7 billion in goodwill (63% of its $7.4B purchase price), and the Confluent acquisition will add more. As IBM shifts from buybacks to M&A, goodwill concentration will continue to increase.

Methodology

Data Sources

This analysis is based on IBM's FY2025 Annual Report (10-K, filed February 24, 2026) and IBM's quarterly reports (10-Q) through September 30, 2025. Financial metrics were extracted using MetricDuck's automated pipeline, which processes SEC filings to calculate ROIC, operating leverage, segment profitability, and other derived metrics. Peer data for ServiceNow (NOW), Mastercard (MA), and Visa (V) was sourced from the same pipeline for consistency. Supplementary data points — including IBM's FY2026 FCF guidance of $15.7 billion and the $12.5 billion AI book of business — were sourced from IBM's Q4 2025 earnings release (8-K) and management commentary.

Limitations

• 10-K MD&A not fully extracted: IBM's 10-K MD&A section had limited extraction coverage. Analysis relied primarily on 10-Q MD&A (38 chunks through Q3 2025), with Q4 figures derived as FY minus 9M. Some Q4-specific management commentary may be absent.
• AI book of business ($12.5B) unverified: This figure appears only in IBM's earnings release, not in SEC filings. It is unclear whether it represents total contract value, annual contract value, or cumulative signings.
• Consulting sub-segmentation unavailable: IBM does not disclose COBOL/legacy modernization revenue separately. The Claude Code threat's specific revenue exposure cannot be quantified — only total consulting's profit contribution (17.1%).
• Confluent impact forward-looking: The Confluent acquisition closed March 17, 2026. Its impact on FY2026 financials is estimated based on public guidance and cannot yet be verified from filed financials.
• Peer fiscal year mismatch: Visa's fiscal year ends September 30 (vs December 31 for IBM, NOW, MA), creating a slight timing mismatch in peer comparisons.
• Post-crash price basis: The valuation analysis uses ~$246 (March 2026 post-crash), not the December 31, 2025 pipeline price. Market price is volatile; multiples shift accordingly.

Disclaimer:

This analysis is for informational purposes only and does not constitute investment advice. The author does not hold positions in IBM, NOW, MA, or V. Past performance and current metrics do not guarantee future results. All data is derived from public SEC filings and may contain errors or omissions from the automated extraction process. Derived calculations are documented with formulas and source inputs throughout.