Trafigura’s Henry Van said hyperscaler plans imply AI related copper use could reach 2.5–3 million metric tons a year in 2028, versus about 1 million in 2026. A secondary summary reports a more cautious market view of about 1.6 million tons in 2028, but the available Bloomberg report does not confirm that figure.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer

Create a landscape editorial hero image for this Studio Global article: What did Trafigura’s Henry Van, Saad Rahim and Graeme Train say during London’s 2026 LME Week, including Bloomberg’s October 20 Base Metals. Article summary: The available reporting supports a distinction between copper demand implied by hyperscalers’ announced AI projects—up to 3 million metric tons annually by 2028—and a more conservative, market-implied figure of about 1.6. Topic tags: general, news, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts wi
Hyperscaler investment plans could mean a steep increase in copper demand from AI infrastructure—but announced projects are not the same as completed data centers. At Bloomberg’s October 20, 2026, Base Metals event in London, Trafigura’s Henry Van said those plans imply annual AI-related copper use of 2.5–3 million metric tons in 2028, up from about 1 million in 2026. 1
The available reporting supports that project-based estimate and a more cautious market view reported elsewhere. It does not provide enough detail to verify every figure or reconstruct a full account of remarks by Van, Saad Rahim, and Graeme Train.
The key distinction is between copper demand implied by announced hyperscaler projects and demand that will materialize as construction proceeds. Bloomberg reported Van’s estimate as an implication of spending plans for large-scale cloud computing and AI data-center infrastructure—not as a guarantee that all planned projects will be built on time. 1
Bloomberg’s event agenda placed AI, data centers, power infrastructure, supply constraints, and trade-policy uncertainty among the issues facing base-metals markets. 2 Those factors help explain why the headline demand number is best read as a buildout scenario rather than a settled forecast.
A secondary summary puts the more conservative market view at about 1.6 million metric tons of AI-related copper demand in 2028. 3
8 That would still be above the roughly 1-million-ton 2026 estimate reported by Bloomberg, but the available Bloomberg article excerpt does not confirm the 1.6-million-ton figure or explain how it was derived.
1
The sources provided also do not establish that Van explicitly called the higher estimate “not our base case.” The safer conclusion is that the high-end number describes demand implied by announced plans, while the lower figure appears in secondary coverage as a more cautious view.
A separate LME Week metals debate offered a snapshot of audience opinion, not a forecast of data-center construction. Fastmarkets reported that copper went from almost three-quarters of the audience vote before the debate to 50% afterward; tin received 26.5%. 7
That change is consistent with a debate over how much of the anticipated AI buildout will translate into near-term demand. It does not, by itself, show why individual attendees changed their views or quantify the risk of project delays.
A Reuters-republished report attributes to Graeme Train the view that copper’s outlook is complex: AI data centers add to demand, but that demand could falter. The excerpt also says Train noted that estimates had been upgraded, but it does not provide enough context to explain the revision or its assumptions. 15
A separate social-media post attributes to Saad Rahim a broader view that demand for key industrial metals could grow at twice the pace of the previous decade. That post does not establish a copper-specific forecast or link the statement directly to AI data centers. 35
The available source excerpts do not substantiate several more detailed claims, including data centers’ current share of copper demand, an additional 1–2 million tons of demand over five years, roughly 900,000 tons of US stockpiling, or about 2 million tons of new supply in 2027. They also do not establish the claim that copper demand would fall in 2026 without data-center construction.
Those figures may require fuller event coverage or transcripts to verify. Based on the material available here, the clearest takeaway is narrower: hyperscaler plans point to a potentially large increase in AI-related copper use by 2028, while the scale and timing of actual demand remain uncertain.
Studio Global AI
This page includes a source-backed answer you can continue inside Studio Global.
Trafigura’s Henry Van said hyperscaler plans imply AI related copper use could reach 2.5–3 million metric tons a year in 2028, versus about 1 million in 2026.
Trafigura’s Henry Van said hyperscaler plans imply AI related copper use could reach 2.5–3 million metric tons a year in 2028, versus about 1 million in 2026. A secondary summary reports a more cautious market view of about 1.6 million tons in 2028, but the available Bloomberg report does not confirm that figure.
At a separate LME Week metals debate, copper’s share of an audience poll fell from almost three quarters before the debate to 50% afterward; tin received 26.5%.
Trafigura’s Henry Van said hyperscaler plans imply AI related copper use could reach 2.5–3 million metric tons a year in 2028, versus about 1 million in 2026. A secondary summary reports a more cautious market view of about 1.6 million tons in 2028, but the available Bloomberg report does not confirm that figure.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer

Create a landscape editorial hero image for this Studio Global article: What did Trafigura’s Henry Van, Saad Rahim and Graeme Train say during London’s 2026 LME Week, including Bloomberg’s October 20 Base Metals. Article summary: The available reporting supports a distinction between copper demand implied by hyperscalers’ announced AI projects—up to 3 million metric tons annually by 2028—and a more conservative, market-implied figure of about 1.6. Topic tags: general, news, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts wi
Hyperscaler investment plans could mean a steep increase in copper demand from AI infrastructure—but announced projects are not the same as completed data centers. At Bloomberg’s October 20, 2026, Base Metals event in London, Trafigura’s Henry Van said those plans imply annual AI-related copper use of 2.5–3 million metric tons in 2028, up from about 1 million in 2026. 1
The available reporting supports that project-based estimate and a more cautious market view reported elsewhere. It does not provide enough detail to verify every figure or reconstruct a full account of remarks by Van, Saad Rahim, and Graeme Train.
The key distinction is between copper demand implied by announced hyperscaler projects and demand that will materialize as construction proceeds. Bloomberg reported Van’s estimate as an implication of spending plans for large-scale cloud computing and AI data-center infrastructure—not as a guarantee that all planned projects will be built on time. 1
Bloomberg’s event agenda placed AI, data centers, power infrastructure, supply constraints, and trade-policy uncertainty among the issues facing base-metals markets. 2 Those factors help explain why the headline demand number is best read as a buildout scenario rather than a settled forecast.
A secondary summary puts the more conservative market view at about 1.6 million metric tons of AI-related copper demand in 2028. 3
8 That would still be above the roughly 1-million-ton 2026 estimate reported by Bloomberg, but the available Bloomberg article excerpt does not confirm the 1.6-million-ton figure or explain how it was derived.
1
The sources provided also do not establish that Van explicitly called the higher estimate “not our base case.” The safer conclusion is that the high-end number describes demand implied by announced plans, while the lower figure appears in secondary coverage as a more cautious view.
A separate LME Week metals debate offered a snapshot of audience opinion, not a forecast of data-center construction. Fastmarkets reported that copper went from almost three-quarters of the audience vote before the debate to 50% afterward; tin received 26.5%. 7
That change is consistent with a debate over how much of the anticipated AI buildout will translate into near-term demand. It does not, by itself, show why individual attendees changed their views or quantify the risk of project delays.
A Reuters-republished report attributes to Graeme Train the view that copper’s outlook is complex: AI data centers add to demand, but that demand could falter. The excerpt also says Train noted that estimates had been upgraded, but it does not provide enough context to explain the revision or its assumptions. 15
A separate social-media post attributes to Saad Rahim a broader view that demand for key industrial metals could grow at twice the pace of the previous decade. That post does not establish a copper-specific forecast or link the statement directly to AI data centers. 35
The available source excerpts do not substantiate several more detailed claims, including data centers’ current share of copper demand, an additional 1–2 million tons of demand over five years, roughly 900,000 tons of US stockpiling, or about 2 million tons of new supply in 2027. They also do not establish the claim that copper demand would fall in 2026 without data-center construction.
Those figures may require fuller event coverage or transcripts to verify. Based on the material available here, the clearest takeaway is narrower: hyperscaler plans point to a potentially large increase in AI-related copper use by 2028, while the scale and timing of actual demand remain uncertain.
Studio Global AI
This page includes a source-backed answer you can continue inside Studio Global.
Trafigura’s Henry Van said hyperscaler plans imply AI related copper use could reach 2.5–3 million metric tons a year in 2028, versus about 1 million in 2026.
Trafigura’s Henry Van said hyperscaler plans imply AI related copper use could reach 2.5–3 million metric tons a year in 2028, versus about 1 million in 2026. A secondary summary reports a more cautious market view of about 1.6 million tons in 2028, but the available Bloomberg report does not confirm that figure.
At a separate LME Week metals debate, copper’s share of an audience poll fell from almost three quarters before the debate to 50% afterward; tin received 26.5%.