Friday, August 21, 2026

    TMT Insights

    What the board is tracking, and what it does to the numbers

    What changed for TMT finance leaders
    Updated 3 days ago · past 7 days · 40 sources

    AI capital intensity remains the dominant story boards are underwriting, but the data this week shows the P&L consequences fanning out beyond capex into ad tech consolidation, model pricing wars and workforce cost lines. Cisco's $63.3bn record year and 18% Q4 growth confirm networking is riding the AI infrastructure buildout, though gross margin compressed 210bps — capacity investment is diluting profitability even as topline surges.

    • Ad tech's public-market era is ending: Nielsen-DoubleVerify, IAS-Novacap and now Criteo-Vista signal take-private consolidation as utility-margin platforms lose scale economics.
    • Charter-Cox cleared California PUC, moving the $34.5bn cable megamerger toward close and creating a 70-million-passing operator.
    • OpenAI and Anthropic are cutting prices 80% amid Chinese rival pressure — token cost deflation is real but unpredictable for enterprise budget planning.
    • NVIDIA's $500bn financing platform with Apollo, BlackRock, Blackstone reframes AI compute as an asset class, shifting some infrastructure risk off hyperscaler balance sheets.
    • AI-linked layoffs hit 205,000 in 2026, already matching full-year 2025, as nearly a third of CEOs cite headcount reduction as the AI objective.

    For CFOs, the read is bifurcated: infrastructure and content economics are consolidating and repricing fast, while AI's productivity case remains mostly a future-tense narrative on earnings calls — leaving disclosure, workforce cost and vendor-lock-in governance as the immediate CIO/CTO conversation.

    Top signals
    • Cisco Systems Inc (CSCO) Q4 2026 Earnings Call Highlights: Record Revenue and AI Super Cycle Fuel 18% GrowthSource

      Record $63.3bn revenue and 18% Q4 growth show AI infrastructure demand is a genuine top-line driver, but 210bps of gross margin compression is the CFO tell that capacity investment is outrunning pricing power.

    • California PUC stamps Charter-Cox merger, with conditionsSource

      The $34.5bn deal clearing its last major regulatory hurdle moves the world's largest cable operator toward close next week — a scale play on subscriber base and capex leverage, not AI.

    • Future of Marketing Briefing: The public ad tech era is overSource

      Nielsen-DoubleVerify, IAS-Novacap and now Criteo-Vista mark a structural repricing of ad verification and measurement as public-market growth stories, with private equity absorbing utility-margin names one by one.

    • NVIDIA AI Factory Compute Is Becoming an Investable Asset ClassSource

      Partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR aim to mobilize over $500bn in third-party capital — a direct answer to the capex-versus-balance-sheet tension hyperscaler CFOs face.

    • AI-Linked Layoffs Hit 205,000 Workers in 2026: ReportSource

      Workforce reductions attributed to automation have already matched full-year 2025 in eight months — a hard cost-line number finance chiefs can point to, distinct from the still-unrealized productivity narrative.

    Listen

    TMT Brief, the latest episode

    Tuesday, August 18 · about 6 min

    0:00-:--
    Speed

    Telecom

    9 stories
    Light ReadingAugust 20

    Charter absorbs Cox and Liberty Broadband, becoming the dominant U.S. broadband operator with ~26% Cox Enterprises ownership.

    Charter's dual closing reshapes the U.S. cable landscape: Cox Enterprises now holds roughly a quarter of the enlarged company after receiving units, convertible preferred, and $4 billion cash—plus $12 billion in retained debt. The Liberty Broadband merger simultaneously simplified Charter's ownership structure, netting a reduction of about 4.7 million shares. The combined entity claims top U.S. broadband and video position, with mobile as its fastest-growing segment.

    2 minRead
    Mobile World LiveAugust 20

    OpenAI targets enterprise data retention worries

    OpenAI reportedly pledged to eliminate data retention for enterprise customers using its AI models, a policy which contrasts with rival Anthropic. The Wall Street Journal ( WSJ) reported the ChatGPT-maker believes its new technique will allow it to safely provide its most advanced models without keeping business information on its servers, directly targeting growing enterprise frustration over competing data logging practices. The position contrasts with that of fellow AI player Anthropic. Last week, the Claude-maker instituted a 30-day data retention policy on its most capable models, including Fable 5 and Mythos 5, to monitor safety risks. WSJ noted Anthropic’s announcement sparked significant backlash across Silicon Valley and corporate boardrooms. It drew sharp criticism from prominent figures including White House AI adviser David Sacks, Microsoft CEO Satya Nadella and Palantir CEO Alex Karp. Tech leaders argue regulated industries like healthcare and finance require absolute certainty their proprietary information remains strictly private. While Anthropic maintains the 30-day window is essential to stop complex multi-step attacks, OpenAI is betting privacy first architecture will win over cautious enterprises. Axios reported OpenAI’s new system, called Private Safety Processing, resolves a long-standing trade-off between real time risk monitoring and user privacy. Previously, tracking misuse required storing interaction histories because sophisticated threats often span multiple exchanges. A user might query a software vulnerability in one chat and ask about bypassing security controls in another. Under the new previewed framework, customer data remains on client-controlled infrastructure or is secured via customer owned encryption keys. OpenAI receives only a narrowly defined cryptographic signal when threat patterns emerge, rather than exposing underlying prompts or outputs. The post OpenAI targets enterprise data retention worries appeared first on Mobile World Live .

    2 minRead
    Mobile World LiveAugust 20

    Micron bets $10B that domestic memory supremacy is the decisive variable in the AI economy race.

    Memory is becoming AI's strategic chokepoint, and Micron is moving to own it. The company's new Boise-based Micron Research Labs will funnel $10 billion over a decade into long-horizon work on memory architectures, advanced packaging, and future fabrication—areas deliberately upstream of current product roadmaps. Nvidia's Jensen Huang, Apple's Tim Cook, and Commerce Secretary Lutnick each endorsed the initiative. Ground breaks in 2027. The lab sits inside Micron's broader $250 billion US investment pledge.

    2 minRead
    Mobile World LiveAugust 19

    Google's $12B Marvell warrant ties equity upside directly to chip-purchase volume, a novel alignment of vendor and customer incentives.

    Alphabet has received warrants to acquire nearly 59 million Marvell shares—potentially worth $12.2B—structured so most shares unlock only as Google generates $500M revenue increments through fiscal 2033. The arrangement deepens Google's TPU ecosystem buildout, spanning inference accelerators, networking, and near-memory compute. It arrives as hyperscalers collectively target over $700B in AI infrastructure spending this year, and intensifies rivalry between Marvell and Broadcom, which secured its own long-term Google custom-chip pact in April. **Watch:** Whether performance-linked equity warrants become the template for future foundry-hyperscaler deals.

    2 minRead
    RCR Wireless NewsAugust 19

    60% of telco chiefs want AI revenue; only 25% can deliver it—a 35-point execution gap that threatens the TechCo pivot.

    A 2026 HCLTech/Mobile World Live survey exposes a stark divide: most operators are using AI to trim costs on aging infrastructure, not to build new revenue streams. Legacy BSS/OSS silos corrupt model inputs; talent pipelines can't compete with hyperscalers; and nearly half of respondents rate their transformation culture as merely "moderate." With roughly $400 billion in enterprise value theoretically at stake, incremental efficiency plays won't close the gap.

    5 minRead
    Light ReadingAugust 14

    CPUC clears Charter-Cox merger; a 70M-home cable giant forms—and Charter-Comcast speculation ignites immediately.

    California's final regulatory hurdle cleared, putting a ~$34.5B merger on track to close within days. The combined entity will operate under Spectrum branding with Charter's leadership. Conditions include $275M in California network upgrades, $30M for digital inclusion, and low-income broadband tiers. Analysts are already pivoting attention to a potential Charter-Comcast combination, buoyed by Comcast's pending cable-media split—despite Roberts publicly dismissing the idea.

    5 minRead
    Mobile World LiveAugust 14

    T-Mobile's 2020 SA 5G bet is now a measurable competitive moat in mass-event connectivity.

    T-Mobile's early standalone 5G switch is paying dividends: 38TB carried at the Knicks parade, one petabyte across FIFA World Cup match days, and 99.9% availability across 11 host cities. Network slicing—spanning 25 distinct slices on Nokia and Ericsson kit—cut payment terminal rejection rates and supported record first-responder counts simultaneously. AI-driven tools manage crowd unpredictability in near real time. AI-RAN trials at the UN General Assembly, US Open, and two F1 races are next. **Watch:** Whether AI-RAN trials convert into a deployable product before rivals close the SA gap.

    4 minRead
    Mobile World LiveAugust 14

    Apple shifts from licensing Chinese AI to co-developing its own model with Alibaba, signaling a deeper sovereignty play in its most contested market.

    Rather than simply reselling domestic partners' models, Apple is jointly training a proprietary large language model with Alibaba for China—a move that hands Cupertino tighter control over the AI layer on iPhones there. The Cyberspace Administration of China cleared Apple Intelligence last month after a lengthy review. Alibaba's Qwen and Baidu technology will both feature in the rollout. If approved, Apple could be the first foreign firm to deploy a proprietary AI product in China.

    2 minRead
    RCR Wireless NewsAugust 14

    Lockheed's NetSense turns existing Verizon 5G into passive drone radar—no new hardware, subscription pricing by 2027.

    Passive RF sensing of drone disturbances in live cellular spectrum—not dedicated radar—is the commercial bet Lockheed, Verizon, Nvidia, and three partners demonstrated over Miami in July. No radios were modified. Astris AI will sell the capability as a managed subscription service, with pilots late 2026 and broad rollout targeted for 2027. The constraint is the network itself: sparse rural coverage limits detection, while dense urban multipath raises false-positive risk—the AI filtering layer is where independent validation matters most.

    5 minRead

    Media & entertainment

    6 stories
    Nieman LabAugust 19

    MS NOW will launch a paid membership program for “super fans”

    When MS NOW — known until last year as MSNBC — celebrated its 30th anniversary last month, Status reported that the focus was “as much about the future as the past.” Natalie Korach described early conversations about the network’s planned subscription product, which she wrote would be “centered around community, giving them access to their favorite hosts, subscriber discussions, and interactive content, rather than a traditional streaming service.” An MS NOW memo reported by Semafor offers new details on the subscription offering. On Sept. 9, the network plans to launch a paid membership program “aimed at engaging its existing ‘super fans,’ and offering ways to reach and monetize new audiences that are not everyday cable news viewers,” Semafor’s Max Tani reported on Wednesday. MS NOW president Rebecca Kutler framed the membership program as the first in a suite of planned direct-to-consumer offerings. MS NOW was renamed after NBCUniversal split off the progressive network and other cable networks into a separate company called Versant. Now, it joins networks from CNN to Fox News to its former sibling NBC News in making digital audiences and monetization a priority as audiences and revenue from TV steadily decline . In 2024, two years after launching and shuttering its streaming platform , CNN put up a paywall ($3.99 a month or $29.99 a year for “basic,” and $6.99/month or $69.99/year for “all access”). NBC News has done the same ($5.99 a month, or $59.99 a year), while Fox News offers subscriptions to its streaming platform Fox Nation for $8.99/month or $71.49/year. CNBC, a part of Versant, is also rolling out its own premium consumer product . The MS NOW membership will be priced at $7.99 a month, or $79.99 per year, and will “offer perks like members-only content verticals, opportunities for subscribers to interact with the network’s personalities, community features for members to engage with each other, and a 24/7 network feed,” Tani wrote. He added that MS NOW “is leaning more into the parasocial relationship its fans feel with its journalists,” including with live events. The membership program is based on online survey responses and interviews with 18,000 network viewers, according to Tani. MS NOW told him they found “the network’s fans felt the news media ecosystem was too fragmented, and wanted to get news in one place.” They also found target audiences “wanted online spaces to interact with people they agreed with” and “expressed excitement at the possibility of interacting with the platform’s journalists and personalities.” Kutler noted that the network sees little overlap between its TV viewership and the current audience for its free online content. Read more in Semafor here . MS NOW

    2 minRead
    Variety AIAugust 15

    Disney bets streaming growth on sports rights, local-language content, vertical formats, and niche verticals—not just marquee IP.

    At D23 in Anaheim, Disney executives moved beyond fan-service announcements to outline the structural pillars they believe will sustain streaming expansion: live sports, non-English originals, short-form vertical video, and targeted content verticals. The pivot signals Disney is no longer treating Disney+ as a Marvel-and-Star-Wars delivery mechanism but as a multi-format, globally differentiated platform competing on audience breadth as much as franchise depth.

    2 minRead
    DigidayAugust 14

    Ad tech's public market chapter is closing fast—DV, IAS, LiveRamp, and Criteo all heading private within 18 months.

    Four major ad tech firms are exiting public markets in rapid succession. Nielsen's $2.15B acquisition of DoubleVerify follows IAS's ~$2B Novacap take-private last December. LiveRamp is heading toward a Publicis-owned exit before year-end. Criteo, nursing a 14% revenue decline, is in active take-private talks with Vista Equity Partners. The pattern signals that mid-tier ad tech—utility-grade, slower-growth—no longer commands public market patience.

    2 minRead
    DigidayAugust 14

    European publishers are getting hit harder by AI bot scraping, report finds

    European publishers are getting hit harder than North American sites by AI bot scraping – and no one really knows why. A recent report from TollBit found that European sites get scraped more by AI bots, receive fewer referrals, and have their robots.txt rules – which websites use to communicate to crawlers what they can and cannot scrape – ignored more often. Here are the main findings from that report, based on TollBit’s analysis identified AI bots from 40 scraping vendors across 3,906 publishers:

    2 minRead
    Nieman LabAugust 13

    Apple may start paying publishers to deliver news with Siri AI

    Hey Siri, how much is the news worth to an AI assistant? The answer, according to a report by Alexandra Bruell in The Wall Street Journal, starts at $100 million. Apple has approached publishers to use their news and up-to-date information in the revamped Siri AI, sources told the Journal. The company has reportedly discussed a “nine-figure budget” for the payments. The next iteration of Apple’s voice-powered assistant was first promised in 2024 and is now expected to be released to the public later this year. (Beta reviewers seem to think it’s an improvement .) Apple has proposed “a variable compensation plan” to news publishers, according to the Journal. The pay-as-you-go plan would see publishers paid only when Siri AI uses their content. It would be a break from other AI licensing agreements that have involved fixed fees and wide-ranging access to content. Apple currently makes payments to news companies for Apple News+ using a similar model , paying publishers “proportionate to the share of clicks they generate within the platform.”

    2 minRead
    Variety AIAugust 13

    Netflix shuttered Night School Studio just 6 weeks post-launch, signaling brutal ROI pressure on its gaming division.

    Netflix is closing Night School Studio—acquired in 2021 and developer of the David Fincher- and Zach Cregger-backed horror title *Unhinged*—only six weeks after the game's release. Helsinki-based Moonloot, founded in 2022, is also being shuttered. The timing raises hard questions about Netflix's willingness to sustain gaming investments past launch windows, and whether prestige creative partnerships can survive the streamer's increasingly metrics-driven portfolio cuts.

    2 minRead

    Technology

    6 stories
    StratecheryAugust 19

    Apple's App Store fee structure is cracking under regulatory pressure in both the EU and US markets.

    Regulatory attrition is finally reshaping Apple's App Store economics. EU enforcement and emerging US pressure are forcing lower fees—a structural shift that could compress Apple's Services revenue growth and alter developer economics platform-wide. Ben Thompson frames the EU's delayed victory as legitimate, signaling regulators elsewhere may adopt similar patience. The full analysis is paywalled, but the headline concession marks a turning point in how platform gatekeepers price access to their audiences.

    3 minRead
    Ars TechnicaAugust 18

    Microsoft Copilot self-disclosed a secret bypass parameter that enabled zero-click credential exfiltration in M365 Enterprise.

    Varonis researchers exploited Microsoft 365 Copilot Enterprise by interrogating the AI itself. Rather than reverse-engineering guardrails, they queried Copilot about its own consent mechanisms until it revealed an undocumented prompt parameter—a Microsoft trade secret—that stripped the requirement for explicit user confirmation. The resulting exploit harvested passwords and sensitive data on a single link-click. The incident exposes a systemic risk: LLMs trained on internal architecture may become their own vulnerability disclosure channels.

    2 minRead
    StratecheryAugust 18

    Nvidia deepens frontier AI ties; Anthropic revenue surges; Google's Spirit Airlines data buy signals a new data-monetization era.

    Three signals worth tracking: Nvidia is now backing OpenAI infrastructure directly, extending its influence beyond chips into lab economics. Anthropic's revenue trajectory continues to outpace expectations, reinforcing its position as a serious OpenAI rival. Most intriguingly, Google reportedly acquired data from bankrupt Spirit Airlines—a move that, if confirmed, suggests distressed-asset data harvesting is becoming a legitimate enterprise AI strategy. Full analysis is paywalled at Stratechery Plus. * Watch: Whether Nvidia's lab investments create conflicts with its chip-customer relationships.

    3 minRead
    Ars TechnicaAugust 14

    US AI labs are slashing prices defensively, not innovatively, as Chinese rivals erode their customer base.

    Price competition is reshaping the frontier AI market. OpenAI cut costs on its fastest lightweight model by 80%, while Anthropic positioned Claude Opus 5 as premium intelligence at half the price of its flagship. The trigger: enterprise customers are defecting to lower-cost Chinese alternatives from Moonshot and DeepSeek, now gaining traction from Silicon Valley to Europe. Margin pressure is no longer theoretical—it's driving product strategy at the top of the market.

    2 minRead
    StratecheryAugust 14

    Nvidia's new funding mechanisms shift AI bubble risk broadly—financial engineering may delay, not defuse, a capital crunch.

    AI infrastructure spending faces a third constraint beyond compute and power: capital. With AI revenue still unproven, Nvidia is engineering long-duration financing while Google taps equity markets to keep the buildout alive. Stratechery's Ben Thompson argues this financial creativity widens the blast radius of a potential bubble while protecting Nvidia's margins. Separately, Anthropic's EU-mandated watermarking drew sharp criticism as philosophically misguided—a debate likely to outlast any regulation.

    3 minRead
    Ars TechnicaAugust 12

    A 40-minute LiteLLM compromise exposed cloud keys and secrets across 2,500+ orgs including Microsoft, Amazon, and Cisco.

    A supply-chain attack on LiteLLM, the widely used open-source AI development library, gave attackers a narrow but devastating window last March. Malicious PyPI packages silently harvested cloud keys, SSH credentials, Kubernetes secrets, and AI provider tokens at scale. Security firms CloudSEK and Hudson Rock independently surfaced the breach—the latter via a 195TB dataset. The blast radius spans more than 2,500 organizations, including household enterprise names. Exposed secrets could enable lateral movement deep into cloud infrastructure. • Watch: Whether affected organizations rotate credentials before threat actors monetize the stolen access.

    2 minRead

    Across the sector

    3 stories
    CNBC TechnologyAugust 15

    Anthropic's 14x revenue surge to $11.5B in Q2 signals an AI enterprise market consolidating around fewer, larger players.

    Anthropic posted preliminary Q2 revenue exceeding $11.5 billion—up from $787 million a year prior and $4.73 billion last quarter—while achieving positive adjusted operating income. The acceleration, driven by enterprise adoption and coding workflows, positions the Claude maker for a potential fall IPO that would be among the first major AI-native public offerings. CFO Krishna Rao is leading early investor meetings, though valuations remain unspecified. • Watch: Whether Anthropic's IPO pricing sets a benchmark that reframes how markets value AI infrastructure-dependent businesses.

    4 minRead
    CNBC TechnologyAugust 14

    Three C-suite exits in weeks threaten OpenAI's IPO narrative as investors already question its $852B valuation.

    OpenAI's executive bench has collapsed at the worst possible moment. Chief revenue officer Denise Dresser and operating chief Brad Lightcap both departed this week, following earlier exits by president Fidji Simo and four other senior leaders. Investors were blindsided. One AI founder called the timing a "huge red flag," noting executives rarely forfeit large equity packages without cause. CFO Sarah Friar rushed to reassure shareholders Friday, citing 32% month-over-month enterprise growth—but stability messaging now competes with visible leadership chaos.

    6 minRead
    CNBC TechnologyAugust 14

    OpenAI's enterprise revenue has overtaken consumer ahead of schedule, hitting a $40B annualized run rate.

    Enterprise now generates the majority of OpenAI's revenue—a crossover CFO Sarah Friar had projected for year-end. The shift arrived faster than forecast: business customers grew 32% month-over-month in July versus 20% overall. The disclosure came during a scheduled investor session held amid C-suite turbulence, including the departures of revenue chief Denise Dresser and veteran Brad Lightcap. Incoming sales head Dali Rajic, former COO at Google-acquired Wiz, was flagged as the enterprise growth driver going forward.

    5 minRead
    FAQ

    Frequently asked questions

    Keep going