Is the $17B IREN figure real Bernstein, or a single-source echo?

IREN is trading near $39.60 on a forward story that keeps circulating as a Bernstein conclusion: a Street-high $100 target, supported by a business that builds toward roughly $17 billion of revenue on about 2 gigawatts of AI capacity by 2030.

Before you pay for that story, trace it. Only part of the chain is on the record. The $100 target is real, documented, and echoed across independent publishers. The per-megawatt economics under it are real. The figure doing the heavy lifting — $17 billion of revenue by 2030 — is not in any primary Bernstein note I can reproduce, and it is not IREN's own disclosed target. This is the difference between a corroborated figure and a single-source echo, and it is worth $0 if you treat the two as the same thing.

The strand that is on the record: $100 and per-megawatt economics

Bernstein cut its IREN price target to $100 on April 28, 2026 — down from $125 — while keeping an Outperform rating and naming the stock its top pick among AI-focused bitcoin miners. That is attributable, and it appears in multiple independent outlets rather than a single shared wire.

Read the cut closely, because it is a haircut, not a hike: Bernstein trimmed the number on a scaled-back mining footprint and a higher share count from recent equity raises, not on retreating AI ambitions. Its 2030 model keeps cloud revenue at roughly $6 billion, adjusted EBITDA near $5 billion, and adjusted EBITDA margins exceeding 80 percent, with the contracted GPU fleet at 150,000 units. That is the discrepancy that matters: Bernstein's own 2030 top-line, on the record, is about $6 billion — roughly a third of the recycled $17 billion figure.

The per-megawatt economics that make the story credible are likewise documented, though the "5-10x" framing circulating is a paraphrase of harder numbers. On the record: Bernstein valued IREN's 3.6 gigawatts of undeveloped Sweetwater and Oklahoma capacity at about $3 million per megawatt, worth roughly $10.8 billion in its sum-of-the-parts. Independent modeling of the deployed fleet produces revenue densities of roughly $8 million to $12 million per gross megawatt. The directional claim — large-multiple economics relative to what IREN pays to equip — survives.

The $17B link: a one-source echo

What does not survive is the specific forward top-line. I could not locate a second attributable publisher, a primary Bernstein record, or an IREN disclosure stating $17 billion of revenue by 2030. The number almost certainly produces itself: the flagship Sweetwater campus is pegged at about 2 gigawatts, and $17 billion is very close to 2,000 megawatts multiplied by the ~$8.5 million per-megawatt revenue density that circulates. That is a reconstruction — someone ran a per-MW multiple across the 2GW headline and attached the result to Bernstein. A reconstruction is not a finding.

The only "$17 billion IREN" figure I can put on the record is a market-value reference, and it is a different measure entirely: one outlet described an IREN market capitalization around $17 billion in January 2026. Market cap is not forward revenue. Neither a market-cap number nor the per-MW reconstruction should be quoted as a Bernstein 2030 top-line that the firm never printed.

What IREN itself has put on paper

IREN does anchor a multi-gigawatt 2030 capacity plan — that part is company-documented, not analyst speculation. It energized the 1.4-gigawatt Sweetwater 1 site in Texas on May 1, 2026, is building the wider Sweetwater campus toward about 2 gigawatts, announced a 1.6-gigawatt Oklahoma campus in February, and reported more than 4.5 gigawatts of secured power that it later framed as roughly 5.8 gigawatts — the largest power position among the listed neocloud cohort. On the compute side it contracted 150,000 GPUs and guided to $3.4 billion of annualized run-rate revenue by the end of 2026.

All of that supports a genuine capacity thesis: a company with 5-plus gigawatts of power, a 2GW flagship, and a funded $25–30 billion fiscal-2027 capex plan is a multi-gigawatt 2030 story on its own documents. What those documents do not support is the $17 billion revenue figure. IREN's own revenue run-rate target is $3.4 billion for end-2026 — structurally several billion below any $17 billion 2030 line, on any aspirational ramp.

Verdict, and the break condition

Split the chain and the conviction sorts itself:

• The $100 Street-high target — filed and corroborated, but it is a cut to $100, and it does not require $17 billion of revenue to be meaningful.
• The per-megawatt economics — documented and directionally large.
• The $17 billion / 2030 top-line — unreplicated. No second publisher, no primary note, no matching IREN disclosure as of publication. It supplied no independent conviction; whatever case IREN has rests on the $6 billion top-line, the $5 billion EBITDA, and the funded gigawatt roadmap — a strong enough thesis that inflating the top-line by roughly three times is worse than unnecessary.

The break condition that would upgrade the figure: a second attributable Bernstein record or an independent outlet carrying the same $17B-by-2030 line, or IREN disclosing a matching multi-year revenue roadmap. None exists today. Until one does, treat "$17 billion by 2030" as a label, not a number — and ask yourself which multiple you are actually paying, one priced against $17 billion or the $6 billion that is really on the record.