KDDI Invents a Profit Measure — and Buries the Fraud Story
A new adjusted-profit bridge, a Kakaku.com exit, and a data-center blueprint that deliberately rejects the megawatt arms race.
9433.T · Earnings Call · 2026-11-06
The beat beneath the miss
KDDI's fiscal Q2 headline looked contradictory: operating revenue and profit both exceeded the initial plan, adjusted operating profit rose 13.9% year on year, yet profit for the period came in below consensus. President Matsuda addressed it head-on: “profit for the period is below consensus, partly due to the defense tax increase and higher interest expenses” — Hiromichi Matsuda, President or CEO · 2026-11-06. The fix was presentational as much as operational. Finance chief Aketa introduced "adjusted profit" — a non-IFRS measure that strips out restructuring and portfolio gains — and walked through the bridge: “the total of adjustments came to JPY 73.8 billion. So IFRS number of JPY 559.3 billion, and we have added JPY 73.8 billion to that” — Kenji Aketa, Executive (Finance/Accounting) · 2026-11-06. The single largest piece was the Kakaku.com remeasurement gain, JPY 34.9 billion, which KDDI removed from its "real" earnings. This is a company-unique communication shift. For a management that has spent years selling steadiness — 20 consecutive terms of profit growth, as one analyst noted — introducing an adjusted metric is a tacit admission that the IFRS line is about to get noisier. The October board resolution to review the portfolio, withdrawing from businesses that fail ROI or viability tests, means more gains and impairments are coming. The Kakaku.com sale alone is expected to bring roughly JPY 110 billion of cash, inclusive of tax effects, which KDDI intends to recycle into a buyback.The LTV machine finally shows its arithmetic
The more substantive story is in Telecom Core. Mobile communication revenue rose JPY 28 billion, and the broader non-mobile piece added JPY 34.3 billion — over half of which Matsuda attributed to cost structure transformation rather than volume. The churn rate fell to 1.18%, down 0.03 points, and mobile ARPU rose 1.6%. This is the payoff from a strategy KDDI has been telegraphing for at least five quarters: stop paying for short-term subscribers, redirect that spend to existing users, and let lifetime value compound. The prior calls show how long this has been building. In November 2025 Matsuda insisted KDDI was “not that we're having a head-on competition with our peers who are pouring promotional expenses” — Hiromichi Matsuda, President · 2025-11-08. By August 2026, Sasaki said promotional-expense suppression was already "a large factor" behind profit growth, and that “there's nothing decided specifically with respect to the next round of price hike” — Masami Sasaki, Executive/Business Unit Head · 2026-08-07 — a signal that the price lever is largely spent. The new element is quantification. Matsuda now says more than half of the non-mobile Telecom Core gain comes from cost control, and he is explicit that the first-half beat will be reinvested: roughly JPY 30 billion of overachievement is being shifted into second-half growth spending and "responses to environmental changes." The analyst pushback is worth noting because it has not gone away. Tokunaga and Masuno both pressed on the second-half profit step-down; Masuno asked whether a 3% core profit increase is really achievable without another price revision. Sasaki's answer was essentially yes — through brand migration from UQ to au, higher-tier plan penetration above 4 million subscribers, and povo. “Povo's growth was greater than what we expected” — Masami Sasaki, Executive (likely in charge of Personal Growth or related segment) · 2026-11-06, he said — a rare admission of upside surprise in a call otherwise framed around discipline.Two data-center models — and a market shorting the first one
The strategic centerpiece is KDDI's bifurcated data-center bet. The data center story is no longer one thing. Matsuda split it into a "connectivity model," inherited from TELEHOUSE, where KDDI layers connectivity on top of facilities near urban demand, and an AI factory model at Sakai, Osaka, where KDDI owns the GPUs, the data, and the power and sells compute as a service.That is a deliberate rejection of the hyperscale arms race — and it lands at an awkward moment. The global tape has spent the last 90 days punishing exactly the companies that chased AI data center capacity: the theme is down roughly 29% over that window, with 58 decliners against 7 advancers. KDDI's answer is to compete on layer value per megawatt and low churn rather than raw capacity. Whether the market rewards that differentiation or lumps KDDI in with the capex-heavy crowd is an open question, but the framing is new and specific. The adjacent Physical AI push is more speculative. KDDI set up ELYZA Robotics, targets JPY 30 billion of R&D, 300 specialists, and 1,000 robots, and signed a capital alliance with newmo, Osaka's largest taxi operator with 1,900 vehicles, to build an autonomous-driving loop. Matsuda's phrase — "operate-in-Sakai, learn-in-Sakai" — is a nice line, but the payoff is explicitly a 2030 story: “by FY '30, we think we will start seeing numbers” — Hiromichi Matsuda, President or CEO · 2026-11-06. Meanwhile, the GPU fleet depreciates fast. Matsuda conceded the tension: “the evolution of GPUs is very fast. So they cannot be used over 10 years or 20 years.” — Hiromichi Matsuda, President or CEO · 2026-11-06Our commitment is not to pursue scale in terms of megawatt, but to layer value on top of facilities, land, and building.
The story that went quiet
The most striking absence is the fictitious transaction scandal. In the February 2026 call it was the only thing anyone asked about: JPY 246 billion of cumulative fake advertising revenue at G-PLAN and BIGLOBE, JPY 33 billion of external outflow, and a devastating profit hit. “the impact on revenue is this amount and the profit impact is minus JPY 50 billion” — Hiromichi Matsuda, President and CEO · 2026-02-06, Matsuda said then. It ranked as the company's number-one keyword that quarter by a wide margin. In the current transcript it is nowhere. The closest echo is a single line about "strengthening group governance and cybersecurity measures." That silence is itself information. Either the investigation has been contained and management no longer needs to litigate it on every call, or the topic has simply been crowded out by the new portfolio and data-center narrative. Either way, the governance overhang that dominated six months ago has vanished from the curated keyword set — a classic case of what management stops talking about mattering as much as what it starts. The M&A question also lingers. Analyst Kikuchi did not mince words:Matsuda's response — restructuring the M&A department, being selective, and trusting divestment to fund future deals — was more candid than defensive. Executive Yoshimura added the guardrail: “we will not overly push ourselves to reach JPY 1 trillion by all means, no matter what” — Kazuyuki Yoshimura, Executive (likely in management or strategy) · 2026-11-06.I cannot think of many successful M&As you've did in the past. So maybe you are not suited to M&A deals.