China showed us where streaming is going. This week confirmed the infrastructure isn't ready for it

Tencent Music finished 2025 with 127.4 million paying subscribers. Of these, 20 million are on a super-premium SVIP tier paying five times the regular rate. Subscription revenue increased by 16 percent to $2.53 billion.
The average revenue per paying user grew 10.2 percent year-over-year.
Those numbers are important.
But what has happened in the weeks since those results were published is even more important.

Three things broke open this week

First: Spotify confirmed the royalty concentration problem is structural, not marginal.

Spotify's Loud and Clear 2026 report, published on March 11, reported $11 billion paid into the music industry in 2025, the highest annual figure in the platform's history. 13,800 artists earned at least $100,000. The 100,000th highest-earning artist made $7,300, up from $350 a decade ago.

Read past the headline. The pro-rata payment model, where all subscription and advertising revenue is pooled centrally and divided based on each artist's share of total platform streams, primarily concentrates the bulk of the $11 billion at the top of the catalog. The platform's own report acknowledged an ongoing struggle with bot-driven playlists and AI-generated content that artificially inflate stream counts, while notably omitting any detailed strategy to address it.

$11 billion distributed through a system that rewards stream volume rather than authentic listening engagement, in an environment where synthetic content is actively distorting the measurement of that volume.

That is not a small problem with the infrastructure.
That is the infrastructure.

Second: the UK's House of Lords declared AI training on creative work without consent a "clear and present danger" to the creative industries.

Published on March 6, the House of Lords Communications and Digital Committee report explicitly called for a licensing-first regime,AI companies must obtain consent before training on copyrighted works, share in the value generated from that use, and provide full transparency about what was used and how. The Music Publishers Association welcomed the report directly, stating: "Their creativity has immense value, and their work must never be seen as fodder for tech platforms to exploit."

The regulatory landscape is now clear across the UK, the EU, and the US, following the US Supreme Court's early-March ruling reaffirming human authorship requirements. Licensing with consent is not a future goal; it is the emerging legal standard. The infrastructure concern is no longer whether AI training on creative works needs a license, but whether copyright infrastructure can implement that licensing at the scale and speed needed by millions of rights holders.

Third: AI-generated tracks are on course to outnumber human-created tracks on major streaming platforms before mid-2026.

Deezer adds about 150,000 tracks daily, with roughly a third being purely AI-created. Every major streaming platform has announced measures, but none has shown the infrastructure needed to handle the problem at its current scale. Spotify's spam filtering and impersonation policies are reactive and do not provide the ongoing, cross-platform behavioral detection system necessary to address the issue effectively.

What the China signal and this week's news say together

TME's SVIP results established the business model: focus on per-subscriber revenue at the premium level, increase conversion through artist-exclusive features, and grow ARPPU faster than subscriber numbers. Universal Music Group has publicly stated that the Western rollout will follow this model. Spotify's Premium Platinum launch in November 2025 is the first clear step.

But the TME results also established something else:
At scale, with 528 million monthly active users and 56 percent of independent releases being AI content, the royalty distribution problem does not get any easier.
It compounds.​

Now layer in this week's data:

Spotify is paying $11 billion through a pro-rata model that structurally concentrates value at the top. A regulatory framework requiring consent-based licensing for AI training that cannot be done manually at the necessary volume. An upload environment moving toward mostly synthetic content that no current platform can manage continuously and at scale. YouTube is withdrawing from Billboard charts, further fragmenting the cross-platform visibility landscape.

Each of these developments raises the financial cost of copyright infrastructure that relies on human-mediated navigation rather than on autonomous, intelligent, real-time pursuit of outcomes.

The infrastructure question this week is unavoidable

Spotify's $11 billion paid out through a system that cannot verify whether the streams driving distribution are genuine is not a success story for rights holders. It illustrates how severe the misallocation problem has become.

The House of Lords report's three principles, authorisation, remuneration, and transparency, cannot be achieved through quarterly batch reconciliation and manual licensing workflows. They require an infrastructure that understands purpose: that can verify rights at the point of use, execute licensing programmatically, distribute accurately in real time, and detect fraud continuously across the entire platform landscape.

TME demonstrated the revenue model.
This week revealed the infrastructure gap.

The NIM ecosystem is designed for the architecture that both demand: an intelligent, autonomous, adaptive copyright infrastructure that pursues accurate attribution and fair compensation as an ongoing process, not a periodic administrative task.

The pool is $11 billion and increasing. The question is how much of it actually reaches the rights holders it belongs to , and whether the infrastructure serving those rights holders is asking them which menu to choose, or understanding what they are trying to protect.