JPMorgan just posted the highest quarterly profit ever recorded by a U.S. bank. The only operational AI productivity claim in the entire earnings call was a 30-to-40 percent job reduction in “some areas” — no function named, no EBIT attached, no follow-up question from any analyst on the call.

That gap between what the financial disclosures say and what the AI narrative promised is what this earnings season was always going to reveal. Now that the first major Q2 calls are complete, the picture is specific enough to act on.

Signal 1: The AI ROI that showed up in Q2 is a financing story

#

JPMorgan posted a quarterly profit of $21.2 billion — the highest ever by a U.S. bank in a single quarter, according to LSEG data (Reuters, July 14, 2026). Markets revenue surged 35%. Investment banking fees climbed 30%, lifted by deals including Alphabet’s $85 billion equity offering and JPMorgan’s lead role on the SpaceX IPO — the largest listing in history.

Goldman Sachs nearly doubled its quarterly profit year-over-year: $6.63 billion versus $3.72 billion in Q2 2025. Equities revenue hit a record, up 72%. Investment banking fees jumped 55% (Reuters, July 14, 2026). Goldman advised on $1.2 trillion in M&A in the first half of 2026 alone. Bank of America reported trading records; it has helped raise nearly $500 billion for AI-related companies since 2025, representing 60% of all AI-related fundraising across investment-grade debt, leveraged finance, and equity capital markets. Wells Fargo beat on trading and loan growth. Citigroup beat profit estimates, though shares fell on expense concerns.

Every bank CEO mentioned AI. The framing was identical across all five calls: AI as a capital formation and deal flow opportunity, not an operational efficiency story.

Goldman CEO David Solomon said it with precision: “The buildout of AI infrastructure remains in its early stages, and we believe this multi-year investment cycle will continue to drive elevated levels of strategic activity, financing, and capital formation across markets” (Reuters, July 14, 2026). Citigroup CEO Jane Fraser told investors AI was “dominating a lot of the conversation” with spending on technology, data centers, energy, and defense accelerating. Bank of America CEO Brian Moynihan cited “AI-driven investments across the board” as a supporting pillar of the economy.

Those are all real observations about where AI capital is flowing. They are also entirely different from the claim that AI is improving these organizations’ own internal operations.

Signal 2: The only operational productivity claim — and what is missing from it

#

JPMorgan’s Jamie Dimon is the only bank CEO who made a concrete operational productivity claim this week. During Tuesday’s earnings call, Dimon said there are areas where the bank has reduced jobs by 30 to 40 percent due to AI, though most affected employees were able to find other positions within the firm (Reuters, July 14, 2026). JPMorgan also disclosed that it has built 1,000 AI use cases across risk functions, marketing, hedging, and note-taking.

Read that claim carefully. It contains: a percentage range, a reassurance about internal redeployment. It does not contain: a named function, a named business unit, a dollar figure, an EBIT attribution, or a line item in the Q2 financial filing. Dimon offered it in approximately two sentences and moved on. No analyst asked a follow-up question.

The contrast with what McKinsey’s State of AI 2025 survey found matters here. McKinsey reported that 39% of organizations said they had seen real EBIT impact from AI (McKinsey Global Survey on the State of AI, November 2025). That figure was self-reported in a survey — organizations answering the question “have you seen EBIT impact?” with yes or no. The bank Q2 earnings calls are a different instrument: financial disclosures prepared under securities law, reviewed by auditors, and subject to investor scrutiny. In that context, operational AI productivity gains that exist and are material are expected to appear somewhere — in the expenses line, the headcount figure, the operational leverage discussion.

What appeared in audited Q2 filings: record deal fees from AI-related capital formation. What did not appear: a documented EBIT improvement from internal AI deployment at any of the five banks. One CEO made a vague operational claim in passing. None attached a number to it that could survive an audit.

Signal 3: IBM — where the AI narrative met the accountability test

#

Not every company reporting this week had a record quarter. IBM shares fell 25% after Chief Executive Arvind Krishna acknowledged the company had “faltered” as customers shifted their spending toward AI servers and storage infrastructure rather than IBM’s traditional technology services and consulting (Financial Times, July 14, 2026).

That is the operational side of the same story the banks are telling from the financial side.

IBM’s position in the AI cycle was: we help enterprises deploy AI. The market’s Q2 read: enterprises are deploying AI by purchasing hardware infrastructure from the GPU-and-server ecosystem — not by paying IBM to integrate and configure it. IBM got compressed in the middle. Above it, the infrastructure providers — hyperscalers, chip makers, GPU manufacturers — are capturing the investment premium that Solomon called an “AI capex super cycle.” Below IBM, efficiency gains from AI deployment are either not yet materializing or are being produced through tools that do not require IBM’s services layer.

IBM is the first major example of a corporate AI narrative failing a quarterly earnings accountability test. The company told a coherent story about its AI opportunity for eighteen months. Q2 revealed that its customers were telling a different story with their budgets. A 25% single-day stock decline is the disclosure that the narrative and the financial reality had diverged.

The bonus data point: CPI and the real wage picture

#

June CPI landed the same morning as the bank earnings, and it materially changes the real wage picture from the July 8 analysis.

The Consumer Price Index fell 0.4 percent on a seasonally adjusted basis in June — the largest one-month decline since April 2020 (BLS, July 14, 2026). Year-over-year, CPI is now 3.5 percent, down from 4.2 percent in May. The driver was energy: gasoline prices fell 9.7 percent in June as oil market dynamics shifted in the context of the U.S.-Iran conflict. Core inflation — all items less food and energy — is now at 2.6 percent year-over-year.

The June jobs report posted nominal wage growth of 3.5 percent year-over-year. Nominal wages at 3.5 percent against inflation at 3.5 percent: for the first time since early 2024, real wage growth is approximately at parity rather than definitively negative. The structural labor market problems from the July 8 analysis are not resolved. The quits rate and hires rate have not moved. But the guaranteed real wage loss that characterized the first half of 2026 has narrowed sharply. For anyone planning a compensation conversation in Q3, this changes the data framing available to them.

Three shifts to make

#

1) When your organization cites “AI ROI” in the next planning cycle, ask which denominator they are using. The banks’ record profits are real and documented. They do not require any employee to be more productive — they require AI infrastructure deals to be financed and advised on. If your organization’s AI ROI claim is rooted in AI-driven deal flow, capital formation, or financing income, that is a legitimate form of ROI. If the claim is rooted in internal operational efficiency gains, ask whether those gains are documented at the EBIT level or exist only in survey self-reporting. Goldman’s Solomon and JPMorgan’s Barnum both described AI demand in indirect, hard-to-attribute terms this week. Barnum’s exact framing: AI infrastructure demand is “like data centers wind up creating demand for plumbers and electricians — you wind up seeing it in sort of slightly non-obvious places.” That is an honest acknowledgment that the operational productivity signal is still diffuse and hard to isolate. Most organizations are in the same position.

2) Dimon’s 30-to-40 percent claim is the current ceiling for what “documented AI productivity” looks like in a financial disclosure. It is a vague range, unattributed to any function, attached to no financial figure, and it passed without a follow-up question from any analyst covering the largest U.S. bank. If the biggest, most sophisticated, most AI-invested financial institution in the world produces that as its operational productivity disclosure, the expectation that your organization’s internal AI champion has rigorous EBIT documentation should be calibrated accordingly. This is not an argument for complacency. It is an argument for the kind of workflow-level proof the June 29 Career Mechanics playbook describes: time saved, error rate reduced, output per unit time, with a comparator. One of those survives an audit. The other ends a conversation at an earnings call.

3) Watch the tech earnings calls over the next two weeks for whether any company provides an operational EBIT attribution for internal AI use. Alphabet, Meta, Microsoft, and Amazon report beginning July 21. Goldman’s Solomon described AI infrastructure as being in its “early stages” — which is the honest version of saying that the productivity returns from the infrastructure being built are not yet visible in financials. If that framing holds across tech earnings too — all AI investment, no operational productivity attribution — Q2 will have produced the clearest data yet on where the AI value cycle actually is. The banks’ record profits from AI deal fees are real. The question heading into week 4 is whether any major tech company can show something different: AI as an internal efficiency gain rather than an external investment opportunity.

The McKinsey 39% EBIT figure is what organizations said when someone asked them in a survey. Q2 earnings is what the same organizations said when the SEC was watching. Goldman made the most money it has made in years. JPMorgan posted the highest quarterly profit ever recorded by a U.S. bank. Not one of them disclosed a documented, attributed, audited EBIT improvement from internal AI deployment. The gap between those two measurement regimes is the most important data point in this earnings season — and it has not received a single headline.

Got a reaction to this read — an earnings transcript I missed, or a Q2 data point that changes the picture? These signals only get sharper when they are tested against what other people are tracking.

Email me at jackson.rodriguez@tlnw.uk with your reaction.

References

#

• Reuters (July 14, 2026). “JPMorgan posts highest quarterly profit ever by a U.S. bank as dealmaking, stock trading surge.” https://www.reuters.com/business/finance/jpmorgan-profit-rises-investment-banking-boom-2026-07-14/ (Accessed July 15, 2026)
• Reuters (July 14, 2026). “Goldman Sachs profit tops estimates on trading boom, corporate deal spree.” https://www.reuters.com/legal/transactional/goldmans-profit-jumps-trading-surge-corporate-deal-spree-2026-07-14/ (Accessed July 15, 2026)
• Reuters (July 14, 2026). “Wall Street banks see AI ‘super cycle’ set to boost deals, financing.” https://www.reuters.com/legal/transactional/wall-street-banks-see-ai-super-cycle-set-boost-deals-financing-2026-07-14/ (Accessed July 15, 2026)
• Reuters (July 14, 2026). “Wall Street bank earnings surge, lifted by trading and investment banking.” https://www.reuters.com/legal/transactional/wall-street-bank-earnings-surge-lifted-by-trading-investment-banking-2026-07-14/ (Accessed July 15, 2026)
• Financial Times (July 14, 2026). “IBM shares plunge 25% as customers shift spending to AI.” https://www.ft.com/content/da478c37-7a32-415d-9f30-3b2981149f95 (Accessed July 15, 2026)
• U.S. Bureau of Labor Statistics (July 14, 2026). “Consumer Price Index — June 2026.” https://www.bls.gov/news.release/cpi.nr0.htm (Accessed July 15, 2026)
• McKinsey Global Institute (November 2025). “The State of AI in 2025: Agents, Innovation, and Transformation.” https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai (Accessed July 15, 2026)