SAMHSA Released Its Addiction Survey Earlier Than Ever. The Number That Matters Isn’t the Headline One.

SAMHSA put out the 2025 National Survey on Drug Use and Health on July 27 — the earliest the agency has ever released its flagship annual count of who in America is using what, and who’s struggling with it. The topline number is the one every outlet will lead with: roughly 45 million Americans age 12 and older, 15.3% of the country, met clinical criteria for a substance use disorder last year. Split it out and it’s 26.0 million with a drug use disorder and 25.7 million with an alcohol use disorder — the two overlap for a lot of people, which is why they don’t neatly add to the 45 million total.

A survey is not a headline number. It’s 45 million individual mornings where someone woke up and the first thing on their mind was a substance, not a person.

Here’s the number that should get more attention than it will: buried in the same release, SAMHSA counted roughly 11,800 people a day starting alcohol use for the first time in 2025, another 11,500 a day trying nicotine vaping for the first time, and about 6,600 a day using marijuana for the first time — somewhere around 30,000 first-time initiations a day, across just those three substances, in a single year. That’s not the 45 million with an existing disorder. That’s the pipeline behind it, and it’s the number that predicts what next year’s survey looks like more than any single-year prevalence figure does.

There’s a genuinely good sign inside the same data: youth and young-adult indicators — major depressive episodes, co-occurring substance use disorder, suicide-related measures — all declined in 2025. Whatever’s driving that (better school-based screening, more accessible teen mental health care, a cultural shift in how younger cohorts relate to drinking and vaping) is worth understanding and replicating, not just noting in passing. It’s one of the only unambiguously encouraging threads in this year’s release.

Read against the CDC’s provisional count of roughly 69,147 overdose deaths for the year ending January 2026 — a 13.2% decline — the picture that emerges isn’t contradiction, it’s two different clocks running at different speeds. Overdose deaths are a lagging indicator, shaped heavily by supply-side factors like fentanyl potency. Initiation and prevalence are closer to a leading one. A treatment provider or facility planning next year’s capacity should be reading the initiation numbers, not just the death-count trendline, because the first is a better preview of who’s going to need a bed in eighteen months than the second is.