Ziff Davis CEO: if OpenAI can find $750 billion for data centers, it can find money for publishers, too

Ziff Davis' CEO challenged the Justice Department's position in an AI copyright case that mandatory publisher licensing could hinder development of a robust U.S. AI industry. The author argues licensing is already operational through deals and new platforms, while OpenAI alone expects to spend $750B on compute through 2030; by comparison, U.S. music royalties are under $20B annually. The commentary contends that compute and energy costs—not content-license fees—are the real barriers to frontier-model competition, and warns that absent licensing publishers may expand paywalls and restrict open-web content.

Analysis

The economically relevant issue is not whether licensing is operationally possible, but whether courts establish a compensable-use standard that converts web content from a zero-cost model input into a recurring COGS line. That outcome would modestly pressure frontier-model gross margins but is unlikely to alter the compute-driven entry barrier; it would instead favor scaled platforms (MSFT, GOOGL, AMZN, META) that can amortize rights costs across distribution, cloud, and consumer products. Smaller model vendors without proprietary distribution or balance-sheet capacity would face the greater relative margin and fundraising discount over the next 6-18 months.

ZD has asymmetric litigation optionality because even a settlement framework could create an incremental licensing revenue category with little associated fulfillment cost. The market should not capitalize that value until there is a court ruling, disclosed settlement economics, or a broader collective-rate mechanism; the Justice Department position is an adverse legal signal, not a damages estimate. A loss could also weaken publishers' negotiating leverage and preserve the current economics of AI extraction, leaving ZD's core valuation largely dependent on its operating execution rather than litigation.

NET is a second-order beneficiary only if crawler controls, authentication, and usage-metering become a standard part of content licensing. Its potential value lies in becoming infrastructure for permissioned machine access rather than in direct royalty participation, but this is presently unproven and likely immaterial to near-term revenue. The contrarian view is that bilateral deals may remain the dominant outcome: large publishers monetize scarcity while the long tail accepts reduced traffic or free exposure, limiting the addressable market for a universal licensing exchange.

Near term, expect legal headlines to create isolated moves in ZD and publisher equities rather than a broad AI de-rating. Over 1-3 months, watch for discovery rulings, motions-to-dismiss outcomes, AI-company disclosures of content commitments, and evidence that publishers are restricting crawler access; these are more actionable than commentary. The thesis is falsified if courts broadly validate training as fair use or if major labs demonstrate that synthetic, licensed, and first-party data substitute for publisher content without measurable model-quality degradation.

Key Decisions for Investors

• Maintain ZD as a small event-driven watch/long only after a favorable dispositive ruling or disclosed settlement; size for binary legal risk and target a 3-6 month catalyst window. Do not underwrite material licensing revenue absent a disclosed rate, eligible corpus, and contract duration.
• Pair trade for a favorable publisher-rights ruling: long ZD versus short a diversified AI-beta proxy such as QQQ, rather than short MSFT or GOOGL outright. The expected effect is relative multiple support for content owners, while the direct cost impact on hyperscalers should be too small to justify a standalone short.
• Keep NET on a 6-18 month infrastructure watch list, not a news-driven buy. Upgrade only if management discloses paid adoption of AI crawler controls, metering, or licensing workflows sufficient to move revenue guidance; otherwise the narrative contribution is unlikely to justify incremental valuation.
• For AI-platform exposure, favor MSFT/GOOGL over smaller frontier-model challengers if licensing obligations become standardized: incumbents can bundle rights costs into existing products, whereas subscale labs face a higher cost of capital. Exit the relative thesis if courts broadly reject publisher claims or if disclosed licensing costs remain de minimis.