=== TAG ===
Business

=== HEADLINE ===
Amazon Just Bet $75 Billion On Both Sides Of The AI War

=== META_DESC ===
Amazon committed $40B to Anthropic and $35B to its own AI division in the same quarter. The largest single tech infrastructure bet in history — on competing horses.

=== DATE ===
April 23, 2026

=== AUTHOR ===
Jane Sterling

=== READ_TIME ===
9-minute read

=== HERO_IMG ===
img/content.png

=== SCRIPT_LABEL ===
Video Script (9 min, ~1,610 words)

=== SCRIPT ===


=== SCRIPT_HTML ===
<span class="scene">SCENE ONE: THE DEAL</span>On Monday, April 20, 2026, Amazon committed up to twenty five BILLION dollars to Anthropic.

That number alone is staggering. But the real story is bigger. Because two months ago, Amazon committed FIFTY billion dollars to OpenAI. Anthropic and OpenAI are direct competitors. They are racing each other to build the most capable AI on the planet. And Amazon is now backing BOTH of them.

Add it up. Seventy five billion dollars. To two rivals. From one company.

Let me unpack what actually happened on Monday.

Amazon agreed to put five billion dollars into Anthropic immediately. Another twenty billion dollars is committed for the future, tied to what the press release calls, quote, certain commercial milestones. That brings Amazon's total stake in Anthropic, going back to its first investment in 2023, to roughly thirty three billion dollars.

In exchange, Anthropic agreed to spend more than ONE HUNDRED BILLION dollars on Amazon Web Services over the next ten years. That spend will be on Trainium, Amazon's custom built AI chip, and on Graviton, Amazon's processor line. The agreement locks in up to five gigawatts of Trainium compute capacity. By the end of THIS year, Anthropic and Amazon plan to bring nearly one full gigawatt of Trainium2 and Trainium3 capacity online.

If those numbers do not mean anything to you yet, here is the translation. Five gigawatts is roughly the power output of FIVE large nuclear reactors. That is the scale of compute Anthropic just locked in to train and serve future versions of Claude. We are no longer talking about server racks. We are talking about industrial scale infrastructure on the same order of magnitude as a national power grid.

So why is this happening NOW.

Anthropic admitted in its own statement that demand for Claude has put what it called inevitable strain on its infrastructure. Their reliability has been wobbling. Their performance has been impacted. The Mythos Preview release earlier this month, which is the most powerful model Anthropic has ever shipped, only made the compute crunch worse. They are running out of room.

OpenAI, meanwhile, has spent the last two months publicly arguing that Anthropic made a strategic mistake by not securing enough compute. OpenAI's chief revenue officer wrote a memo to investors claiming Anthropic was, quote, behind on infrastructure. That memo went public. The pressure on Anthropic to respond was enormous. This deal IS the response.

For Amazon, this is just as strategic. Amazon Web Services is locked in a brutal three way fight with Microsoft Azure and Google Cloud for the future of AI workloads. By backing both Anthropic and OpenAI with massive cash and tying them to AWS infrastructure, Amazon is positioning itself as the neutral compute layer of the AI economy. Whoever wins the model race, Amazon collects rent on the data center.

There is one more wrinkle. OpenAI has now been openly hostile to Microsoft, its long term investor and cloud partner. OpenAI's leadership has said in writing that Microsoft has, quote, limited our ability to reach customers. OpenAI is already using Amazon as a release valve. In the span of two months, Amazon went from being OpenAI's secondary cloud partner to one of its largest investors AND one of Anthropic's largest investors. Amazon is now the bank for both sides of the AI war.

That is the deal. Now let me show you the numbers underneath it.

<span class="scene">SCENE TWO: THE NUMBERS</span>Let me start with the valuation, because it is genuinely insane.

Anthropic is currently valued at three hundred and EIGHTY billion dollars. That valuation came out of a Series G round in February of this year, when the company raised thirty billion dollars in a single round. To put that in context, three years ago Anthropic was valued at around fifteen billion dollars. The company has appreciated more than twenty five times in three years.

And THAT is the conservative number. Reporting from TechCrunch in mid April said Anthropic was actually shrugging off venture capital offers that valued the company at MORE than eight hundred billion dollars. They are turning down funding at twice the price they took it at two months ago.

Now look at revenue. Anthropic announced earlier this month that its annualized revenue had crossed thirty billion dollars. That puts Anthropic AHEAD of OpenAI's reported twenty five billion dollars in annualized revenue for the same period. Let me say that again. Anthropic, the smaller, scrappier, supposedly safety obsessed lab, is now generating more revenue than OpenAI.

OpenAI immediately fired back. Their chief revenue officer claimed that Anthropic's revenue figure is inflated by around eight billion dollars due to accounting choices that make the run rate look bigger than it really is. Anthropic has not publicly responded to that claim. Take that food fight for what it is worth.

Now the compute numbers, which are the real story.

Project Rainier is the codename for the joint Amazon and Anthropic compute build. It already runs on more than one MILLION Trainium2 chips. With this expanded deal, Project Rainier is on track to become one of the single largest AI training clusters on the planet. The five gigawatt commitment will roll out over multiple years. Industry analysts are calling the combined infrastructure across Amazon and Anthropic an eight point five gigawatt buildout when you include planned expansions.

For comparison. The largest single data center in the United States today consumes roughly five hundred megawatts. Five gigawatts is TEN of those, just for one company. Anthropic is committing to operate at the scale of a small country's power consumption.

The chip story matters too. Trainium2 is Amazon's current generation custom AI accelerator. Trainium3, which Amazon previewed at AWS re:Invent late last year, is meant to deliver a meaningful performance jump. Trainium4 is in development. The deal explicitly covers all of these generations, and Anthropic gets first access. That is a competitive moat against any other AI lab that ends up routing through AWS.

For developers, the practical consequence is simple. Claude is going to get more reliable. Mythos Preview, currently locked behind an invitation only program called Project Glasswing, will become more widely available faster as compute comes online. The Opus and Sonnet tiers should see latency improvements in the next two quarters. And pricing pressure on the Claude API will likely ease, because Anthropic now has a guaranteed compute pipeline at preferential rates.

The market reaction was telling. Amazon stock barely moved on the announcement. Analysts seemed to view the deal as priced in. Wall Street has accepted that Amazon is buying its way into AI relevance, and the only real question now is whether the bet pays off before the cash runs out.

<span class="scene">SCENE THREE: THE REAL STORY</span>Now let me tell you what most of the coverage is missing.

This deal is what Wall Street calls circular finance. Amazon gives Anthropic billions of dollars in cash. Anthropic turns around and gives most of that cash back to Amazon by buying AWS compute. Amazon books that compute spend as revenue. Anthropic books the compute as a cost. Both companies look bigger on paper. The actual NEW money in the system is a fraction of the headline number.

This is not unique to Amazon. Microsoft does the same thing with OpenAI. Google does the same thing with Anthropic, separately. Nvidia does it with basically everyone, investing in AI startups that then turn around and buy GPUs from Nvidia. The entire AI economy right now is held together by these recursive deals where the same dollars get counted multiple times across multiple income statements.

Some critics, including a thoughtful piece in HumAI Blog this week, have called this whole structure circular finance and warned that it could be obscuring the true financial strength of the AI industry. The argument is straightforward. If you take out the cloud provider investments, what are the actual unit economics of running Claude or ChatGPT. We do not really know. Nobody is publishing that number.

Hacker News users were sharper. The top comment thread on the Anthropic announcement asked whether Anthropic and OpenAI are racing to IPO precisely so they can convert these paper valuations into real cash before token costs and compute bottlenecks catch up to them. That is a question worth asking. Both companies have explored IPO timelines in the last six months. Both are operating with valuations that assume the current growth curve continues uninterrupted.

There is also the question of conflicts of interest. AWS chief executive Matt Garman gave an interview to TechCrunch on April 8 where he was asked, point blank, whether investing in two direct competitors creates an awkward position. He said it is, quote, an OK conflict because Amazon's job is to support whichever AI lab its customers want to use. That is a defensible position, but it requires AWS to operate a genuine firewall between the two relationships. It is not yet clear that such a firewall exists in practice.

And then there is the geopolitical layer. The White House National AI Framework, released earlier this year, raised explicit concerns about market concentration in AI infrastructure. A scenario where two cloud providers, Amazon and Microsoft, together hold ownership stakes in BOTH leading AI labs and provide the only viable production scale infrastructure for them, is exactly the kind of concentration that policymakers have been warning about. Expect that conversation to get LOUDER.

So where does this leave us.

Anthropic now has the compute it needs to compete head to head with OpenAI for at least the next three years. Amazon has positioned itself as the indispensable infrastructure player regardless of which AI lab pulls ahead. Microsoft and Google are going to have to respond, and quickly, or watch AWS pull away. Developers and enterprises get a more reliable Claude, faster access to future Mythos class models, and likely lower prices.

But somewhere underneath all of it, the question Hacker News asked is the right one. WHO is actually paying for all of this. And what happens when the music stops.

Stay sharp.

Jane Sterling, Sterling Intelligence.

=== ANNOTATED_LABEL ===
Annotated Script (with b-roll &amp; cut cues)

=== ANNOTATED_HTML ===
<span class="scene">SCENE ONE: THE DEAL</span>On Monday, April 20, 2026, Amazon committed up to twenty five BILLION dollars to Anthropic.

That number alone is staggering. But the real story is bigger. Because two months ago, Amazon committed FIFTY billion dollars to OpenAI. Anthropic and OpenAI are direct competitors. They are racing each other to build the most capable AI on the planet. And Amazon is now backing BOTH of them.

Add it up. Seventy five billion dollars. To two rivals. From one company.

Let me unpack what actually happened on Monday.

Amazon agreed to put five billion dollars into Anthropic immediately. Another twenty billion dollars is committed for the future, tied to what the press release calls, quote, certain commercial milestones. That brings Amazon's total stake in Anthropic, going back to its first investment in 2023, to roughly thirty three billion dollars.

In exchange, Anthropic agreed to spend more than ONE HUNDRED BILLION dollars on Amazon Web Services over the next ten years. That spend will be on Trainium, Amazon's custom built AI chip, and on Graviton, Amazon's processor line. The agreement locks in up to five gigawatts of Trainium compute capacity. By the end of THIS year, Anthropic and Amazon plan to bring nearly one full gigawatt of Trainium2 and Trainium3 capacity online.

If those numbers do not mean anything to you yet, here is the translation. Five gigawatts is roughly the power output of FIVE large nuclear reactors. That is the scale of compute Anthropic just locked in to train and serve future versions of Claude. We are no longer talking about server racks. We are talking about industrial scale infrastructure on the same order of magnitude as a national power grid.

So why is this happening NOW.

Anthropic admitted in its own statement that demand for Claude has put what it called inevitable strain on its infrastructure. Their reliability has been wobbling. Their performance has been impacted. The Mythos Preview release earlier this month, which is the most powerful model Anthropic has ever shipped, only made the compute crunch worse. They are running out of room.

OpenAI, meanwhile, has spent the last two months publicly arguing that Anthropic made a strategic mistake by not securing enough compute. OpenAI's chief revenue officer wrote a memo to investors claiming Anthropic was, quote, behind on infrastructure. That memo went public. The pressure on Anthropic to respond was enormous. This deal IS the response.

For Amazon, this is just as strategic. Amazon Web Services is locked in a brutal three way fight with Microsoft Azure and Google Cloud for the future of AI workloads. By backing both Anthropic and OpenAI with massive cash and tying them to AWS infrastructure, Amazon is positioning itself as the neutral compute layer of the AI economy. Whoever wins the model race, Amazon collects rent on the data center.

There is one more wrinkle. OpenAI has now been openly hostile to Microsoft, its long term investor and cloud partner. OpenAI's leadership has said in writing that Microsoft has, quote, limited our ability to reach customers. OpenAI is already using Amazon as a release valve. In the span of two months, Amazon went from being OpenAI's secondary cloud partner to one of its largest investors AND one of Anthropic's largest investors. Amazon is now the bank for both sides of the AI war.

That is the deal. Now let me show you the numbers underneath it.

<span class="scene">SCENE TWO: THE NUMBERS</span>Let me start with the valuation, because it is genuinely insane.

Anthropic is currently valued at three hundred and EIGHTY billion dollars. That valuation came out of a Series G round in February of this year, when the company raised thirty billion dollars in a single round. To put that in context, three years ago Anthropic was valued at around fifteen billion dollars. The company has appreciated more than twenty five times in three years.

And THAT is the conservative number. Reporting from TechCrunch in mid April said Anthropic was actually shrugging off venture capital offers that valued the company at MORE than eight hundred billion dollars. They are turning down funding at twice the price they took it at two months ago.

Now look at revenue. Anthropic announced earlier this month that its annualized revenue had crossed thirty billion dollars. That puts Anthropic AHEAD of OpenAI's reported twenty five billion dollars in annualized revenue for the same period. Let me say that again. Anthropic, the smaller, scrappier, supposedly safety obsessed lab, is now generating more revenue than OpenAI.

OpenAI immediately fired back. Their chief revenue officer claimed that Anthropic's revenue figure is inflated by around eight billion dollars due to accounting choices that make the run rate look bigger than it really is. Anthropic has not publicly responded to that claim. Take that food fight for what it is worth.

Now the compute numbers, which are the real story.

Project Rainier is the codename for the joint Amazon and Anthropic compute build. It already runs on more than one MILLION Trainium2 chips. With this expanded deal, Project Rainier is on track to become one of the single largest AI training clusters on the planet. The five gigawatt commitment will roll out over multiple years. Industry analysts are calling the combined infrastructure across Amazon and Anthropic an eight point five gigawatt buildout when you include planned expansions.

For comparison. The largest single data center in the United States today consumes roughly five hundred megawatts. Five gigawatts is TEN of those, just for one company. Anthropic is committing to operate at the scale of a small country's power consumption.

The chip story matters too. Trainium2 is Amazon's current generation custom AI accelerator. Trainium3, which Amazon previewed at AWS re:Invent late last year, is meant to deliver a meaningful performance jump. Trainium4 is in development. The deal explicitly covers all of these generations, and Anthropic gets first access. That is a competitive moat against any other AI lab that ends up routing through AWS.

For developers, the practical consequence is simple. Claude is going to get more reliable. Mythos Preview, currently locked behind an invitation only program called Project Glasswing, will become more widely available faster as compute comes online. The Opus and Sonnet tiers should see latency improvements in the next two quarters. And pricing pressure on the Claude API will likely ease, because Anthropic now has a guaranteed compute pipeline at preferential rates.

The market reaction was telling. Amazon stock barely moved on the announcement. Analysts seemed to view the deal as priced in. Wall Street has accepted that Amazon is buying its way into AI relevance, and the only real question now is whether the bet pays off before the cash runs out.

<span class="scene">SCENE THREE: THE REAL STORY</span>Now let me tell you what most of the coverage is missing.

This deal is what Wall Street calls circular finance. Amazon gives Anthropic billions of dollars in cash. Anthropic turns around and gives most of that cash back to Amazon by buying AWS compute. Amazon books that compute spend as revenue. Anthropic books the compute as a cost. Both companies look bigger on paper. The actual NEW money in the system is a fraction of the headline number.

This is not unique to Amazon. Microsoft does the same thing with OpenAI. Google does the same thing with Anthropic, separately. Nvidia does it with basically everyone, investing in AI startups that then turn around and buy GPUs from Nvidia. The entire AI economy right now is held together by these recursive deals where the same dollars get counted multiple times across multiple income statements.

Some critics, including a thoughtful piece in HumAI Blog this week, have called this whole structure circular finance and warned that it could be obscuring the true financial strength of the AI industry. The argument is straightforward. If you take out the cloud provider investments, what are the actual unit economics of running Claude or ChatGPT. We do not really know. Nobody is publishing that number.

Hacker News users were sharper. The top comment thread on the Anthropic announcement asked whether Anthropic and OpenAI are racing to IPO precisely so they can convert these paper valuations into real cash before token costs and compute bottlenecks catch up to them. That is a question worth asking. Both companies have explored IPO timelines in the last six months. Both are operating with valuations that assume the current growth curve continues uninterrupted.

There is also the question of conflicts of interest. AWS chief executive Matt Garman gave an interview to TechCrunch on April 8 where he was asked, point blank, whether investing in two direct competitors creates an awkward position. He said it is, quote, an OK conflict because Amazon's job is to support whichever AI lab its customers want to use. That is a defensible position, but it requires AWS to operate a genuine firewall between the two relationships. It is not yet clear that such a firewall exists in practice.

And then there is the geopolitical layer. The White House National AI Framework, released earlier this year, raised explicit concerns about market concentration in AI infrastructure. A scenario where two cloud providers, Amazon and Microsoft, together hold ownership stakes in BOTH leading AI labs and provide the only viable production scale infrastructure for them, is exactly the kind of concentration that policymakers have been warning about. Expect that conversation to get LOUDER.

So where does this leave us.

Anthropic now has the compute it needs to compete head to head with OpenAI for at least the next three years. Amazon has positioned itself as the indispensable infrastructure player regardless of which AI lab pulls ahead. Microsoft and Google are going to have to respond, and quickly, or watch AWS pull away. Developers and enterprises get a more reliable Claude, faster access to future Mythos class models, and likely lower prices.

But somewhere underneath all of it, the question Hacker News asked is the right one. WHO is actually paying for all of this. And what happens when the music stops.

Stay sharp.

Jane Sterling, Sterling Intelligence.

=== ARTICLE_HTML ===


=== YOUTUBE_DESC ===
Amazon just dropped another $25 BILLION on Anthropic — bringing its total stake to roughly $33B and its combined AI bets to $75 billion across both Anthropic AND OpenAI. We break down the numbers, the strategy, and the question Wall Street isn't asking.

On April 20, 2026, Amazon and Anthropic announced an expanded strategic collaboration: $5 billion now, up to $20 billion more tied to commercial milestones, and a $100 BILLION commitment from Anthropic to spend on AWS over the next decade. The deal locks in up to 5 gigawatts of Trainium compute capacity — roughly the power output of five nuclear reactors — to train and serve future versions of Claude.

The context: Anthropic is now valued at $380B (up from $15B three years ago) and reportedly turning down VC offers at $800B+. Annualized revenue just crossed $30B, ahead of OpenAI's $25B (a number OpenAI's CRO publicly disputes as "inflated by $8B"). Amazon also invested up to $50B in OpenAI just two months ago, after OpenAI complained that Microsoft has "limited our ability" to reach Bedrock customers.

Project Rainier — the joint Amazon-Anthropic compute build — already runs on more than 1 million Trainium2 chips and is on track to become one of the single largest AI training clusters on the planet.

In this episode, Jane Sterling breaks down the deal mechanics, the cloud war between AWS, Azure, and Google Cloud, the "circular finance" loop that critics (including Hacker News commenters) say is propping up the entire AI economy, the antitrust risk under the new White House National AI Framework, and what all of it means for developers building on Claude.

⏱ Timestamps
00:00 Scene One — The Deal
03:00 Scene Two — The Numbers
06:00 Scene Three — The Real Story

🔔 Subscribe to Sterling Intelligence for weekly AI coverage that cuts through the hype.
https://www.youtube.com/@SterlingIntelligence

No hype. No filler. Just the signal.

— Jane Sterling, Sterling Intelligence

#Amazon #Anthropic #AWS #Claude #OpenAI #ProjectRainier #Trainium #AINews #AICloud #AIEconomy #SterlingIntelligence #JaneSterling #AIRace #BigTech #AIInvestment

=== TITLES_HTML ===
<li>
      <div class="title-label">Top Pick</div>
      <div class="title-text">Amazon Just Bet $75 BILLION On Both Sides Of AI<span class="title-char">47 chars</span></div>
      <div class="title-reason">Leads with the headline number that makes people stop scrolling, frames it as a "both sides" bet which creates immediate curiosity, and uses CAPS on $75 BILLION for thumbnail-friendly emphasis.</div>
    </li>
    <li>
      <div class="title-label">Alternate 1</div>
      <div class="title-text">Amazon Owns Anthropic Now. Here's The $100B Trick.<span class="title-char">50 chars</span></div>
      <div class="title-reason">Provocative claim ("owns") opens curiosity gap, "$100B trick" hints at the circular-finance angle that's the actual story under the headline. Strong CTR for finance/business viewers.</div>
    </li>
    <li>
      <div class="title-label">Alternate 2</div>
      <div class="title-text">$33 Billion. 5 Gigawatts. The Anthropic Deal Explained.<span class="title-char">55 chars</span></div>
      <div class="title-reason">Two specific, dramatic numbers anchor the title (cumulative investment + compute scale). Appeals to the AI infrastructure / data center crowd who track Trainium and Project Rainier.</div>
    </li>

=== KEYWORDS ===
Amazon, Anthropic, AWS, Claude, Claude Mythos, OpenAI, Project Rainier, Trainium, Trainium2, Trainium3, Graviton, Sam Altman, Dario Amodei, Matt Garman, AI investment, AI cloud war, Microsoft Azure, Google Cloud, Nvidia, AI infrastructure, AI data centers, circular finance, AI bubble, AI economy, AI news 2026, Sterling Intelligence, Jane Sterling, AI weekly, GPT-5.5, AI funding

=== THUMBNAIL_HTML ===
<div class="thumb-spec">
    <h3>Jane's Appearance &amp; Framing</h3>
    <p><strong>Expression.</strong> Calculating, slightly amused. Eyebrow lifted on her left side, mouth in a flat line with the faintest curl at one corner. Not shocked, not laughing. The face you make when you've just done the math on someone else's deal and realized what they actually bought.</p>
    <p><strong>Head position.</strong> Square to camera, chin level, very slight forward lean. Communicates authority, not surprise. The story is the money, not her reaction.</p>
    <p><strong>Wardrobe.</strong> Dark structured blazer, no jewelry. Sterling Intelligence house style. The visual gravity stays on the numbers and the face, not the clothes.</p>
    <p><strong>Eye direction.</strong> Direct to camera, locked. This is a "let me tell you what this actually means" stare. Alternate take: eyes flicked toward the dollar overlay, selling the "look at this number" read.</p>
    <p><strong>Lighting.</strong> Hard key light from upper-right, deep shadow on her left side of the face. Color temp around 4700K. Subtle warm rim light from behind on hair and shoulder to separate from background. Mood is boardroom, not studio.</p>
    <p><strong>Scene setup.</strong> Background is dark charcoal with a faint orange-amber glow far behind her right shoulder (subtle nod to Amazon's brand without screaming it). Shallow depth of field. Optional: a ghosted Amazon smile logo and an Anthropic "A" mark stacked vertically at ~12% opacity behind her left shoulder, suggesting the two companies bracketing her.</p>
  </div>

  <div class="thumb-option">
    <div class="thumb-rank">Option 1 — Best (Money Angle)</div>
    <div class="thumb-text">$75 BILLION</div>
    <p><strong>Position.</strong> Bottom-third, left-aligned, large block stacked on two lines ("$75" on top, "BILLION" below).</p>
    <p><strong>Font.</strong> Bebas Neue Bold or Impact, all caps, tight tracking, slight italic skew of 4 degrees for forward energy.</p>
    <p><strong>Color scheme.</strong> "$75" in gold (#c8a84b), 110% size with subtle outer glow. "BILLION" in pure white (#ffffff), 80% of "$75" size. 3px black stroke around both blocks for legibility against any background.</p>
    <p><strong>Accent detail.</strong> Small red tag below: "ON BOTH SIDES" in Inter Bold, 18px, #dc2626 with 2px white stroke. The red contrasts the gold and signals the conflict-of-interest story the title sets up.</p>
  </div>

  <div class="thumb-option">
    <div class="thumb-rank">Option 2 — Conflict Angle</div>
    <div class="thumb-text">BOTH SIDES</div>
    <p><strong>Position.</strong> Centered upper-third. Large.</p>
    <p><strong>Font.</strong> Inter Black or Montserrat Black, all caps.</p>
    <p><strong>Color scheme.</strong> "BOTH" in white, "SIDES" in red (#dc2626) with white 3px stroke. Slight 8 degree downward angle on the second word for visual tension.</p>
    <p><strong>Accent detail.</strong> Below the main text, two small logos: Anthropic "A" mark on the left, OpenAI flower mark on the right, with a glowing Amazon arrow connecting them. Reads as a financial diagram in a single glance.</p>
  </div>

  <div class="thumb-option">
    <div class="thumb-rank">Option 3 — Compute Scale Angle</div>
    <div class="thumb-text">5 GIGAWATTS</div>
    <p><strong>Position.</strong> Right side, vertically stacked scoreboard style. "5" oversized, "GIGAWATTS" below in smaller caps.</p>
    <p><strong>Font.</strong> JetBrains Mono Bold for the numeral (reads as data / spec sheet), Inter Bold for the unit label.</p>
    <p><strong>Color scheme.</strong> "5" in pure white with a soft red glow, "GIGAWATTS" in muted gray. Small gold underline at the base.</p>
    <p><strong>Accent detail.</strong> Header strip across the top: "PROJECT RAINIER" in 11px small caps, gold. Reads as a credible infrastructure card, not clickbait. Best for the AI infra audience specifically.</p>
  </div>

=== SOURCES_HTML ===
<div class="source-group">
    <h3>Official — Amazon &amp; Anthropic</h3>
    <ul class="source-list">
      <li>
        <a href="https://www.anthropic.com/news/anthropic-amazon-compute" target="_blank" rel="noopener">Anthropic and Amazon expand collaboration for up to 5 gigawatts of Trainium compute</a>
        <span class="src-meta">Anthropic &nbsp;·&nbsp; April 20, 2026</span>
      </li>
      <li>
        <a href="https://www.aboutamazon.com/news/company-news/amazon-invests-additional-5-billion-anthropic-ai" target="_blank" rel="noopener">Amazon and Anthropic expand strategic collaboration</a>
        <span class="src-meta">About Amazon &nbsp;·&nbsp; April 20, 2026</span>
      </li>
    </ul>
  </div>

  <div class="source-group">
    <h3>Media Coverage</h3>
    <ul class="source-list">
      <li>
        <a href="https://www.cnbc.com/2026/04/20/amazon-invest-up-to-25-billion-in-anthropic-part-of-ai-infrastructure.html" target="_blank" rel="noopener">Amazon to invest up to another $25 billion in Anthropic as part of AI infrastructure deal</a>
        <span class="src-meta">CNBC &nbsp;·&nbsp; April 20, 2026</span>
      </li>
      <li>
        <a href="https://www.bloomberg.com/news/articles/2026-04-20/amazon-to-invest-an-additional-5-billion-in-anthropic" target="_blank" rel="noopener">Amazon Invests Additional $5 Billion in Anthropic to Deepen AI Partnership</a>
        <span class="src-meta">Bloomberg &nbsp;·&nbsp; April 20, 2026</span>
      </li>
      <li>
        <a href="https://techcrunch.com/2026/04/20/anthropic-takes-5b-from-amazon-and-pledges-100b-in-cloud-spending-in-return/" target="_blank" rel="noopener">Anthropic takes $5B from Amazon and pledges $100B in cloud spending in return</a>
        <span class="src-meta">TechCrunch &nbsp;·&nbsp; April 20, 2026</span>
      </li>
      <li>
        <a href="https://www.engadget.com/ai/amazon-will-invest-up-to-25-billion-in-anthropic-in-a-broad-deal-225239302.html" target="_blank" rel="noopener">Amazon will invest up to $25 billion in Anthropic in a broad deal</a>
        <span class="src-meta">Engadget &nbsp;·&nbsp; April 20, 2026</span>
      </li>
      <li>
        <a href="https://www.geekwire.com/2026/amazon-doubles-down-on-anthropic-with-25b-investment-mirroring-its-openai-cloud-deal/" target="_blank" rel="noopener">Amazon doubles down on Anthropic with $25B investment, mirroring its OpenAI cloud deal</a>
        <span class="src-meta">GeekWire &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://www.baltimoresun.com/2026/04/21/amazon-anthropic/" target="_blank" rel="noopener">AI startup Anthropic commits $100 billion to Amazon's AWS over next 10 years</a>
        <span class="src-meta">Baltimore Sun &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://www.axios.com/2026/04/21/anthropic-amazon-compute-wars" target="_blank" rel="noopener">Anthropic bites back in the compute wars with Amazon partnership</a>
        <span class="src-meta">Axios &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://www.pymnts.com/artificial-intelligence-2/2026/amazon-and-anthropic-deepen-ties-with-investment-and-hardware-pact/" target="_blank" rel="noopener">Amazon and Anthropic Deepen Ties With Investment and Hardware Pact</a>
        <span class="src-meta">PYMNTS &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://finance.yahoo.com/sectors/technology/articles/amazon-agrees-invest-25-billion-162152225.html" target="_blank" rel="noopener">Amazon Agrees To Invest Up to $25 Billion in Anthropic In Latest Big Tech AI Tie-Up</a>
        <span class="src-meta">Yahoo Finance &nbsp;·&nbsp; April 21, 2026</span>
      </li>
    </ul>
  </div>

  <div class="source-group">
    <h3>Analyst &amp; Independent</h3>
    <ul class="source-list">
      <li>
        <a href="https://www.humai.blog/amazon-just-bet-75-billion-on-two-rivals-only-one-of-them-can-win/" target="_blank" rel="noopener">Amazon's $75B Bet on OpenAI and Anthropic Is Circular Finance</a>
        <span class="src-meta">HumAI Blog &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://www.constellationr.com/insights/news/anthropic-amazon-links-tighten-anthropic-spend-100-billion-over-10-years-aws-capacity" target="_blank" rel="noopener">Anthropic, Amazon links tighten: Anthropic to spend $100 billion over 10 years on AWS capacity</a>
        <span class="src-meta">Constellation Research &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://nerdleveltech.com/amazon-anthropic-100-billion-aws-trainium-5gw-deal" target="_blank" rel="noopener">Amazon-Anthropic $100B Deal: 5GW of AWS Trainium Compute</a>
        <span class="src-meta">Nerd Level Tech &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://securityonline.info/anthropic-amazon-100-billion-deal-5gw-supercomputing-rainier/" target="_blank" rel="noopener">The 100-Billion-Dollar Power Play: Anthropic and Amazon Forge an 8.5GW AI Empire</a>
        <span class="src-meta">Security Online &nbsp;·&nbsp; April 21, 2026</span>
      </li>
      <li>
        <a href="https://www.1950.ai/post/project-rainier-and-the-rise-of-ai-megaclusters-how-amazon-and-anthropic-are-rewriting-compute-at-s" target="_blank" rel="noopener">Project Rainier and the Rise of AI Megaclusters</a>
        <span class="src-meta">1950.ai &nbsp;·&nbsp; April 22, 2026</span>
      </li>
      <li>
        <a href="https://techcrunch.com/2026/04/15/anthropic-shrugs-off-vc-funding-offers-valuing-it-at-800b-for-now/" target="_blank" rel="noopener">Anthropic shrugs off VC funding offers valuing it at $800B+, for now</a>
        <span class="src-meta">TechCrunch &nbsp;·&nbsp; April 15, 2026</span>
      </li>
      <li>
        <a href="https://techcrunch.com/2026/04/08/aws-boss-explains-why-investing-billions-in-both-anthropic-and-openai-is-an-ok-conflict/" target="_blank" rel="noopener">AWS boss explains why investing billions in both Anthropic and OpenAI is an OK conflict</a>
        <span class="src-meta">TechCrunch &nbsp;·&nbsp; April 8, 2026</span>
      </li>
    </ul>
  </div>

  <div class="source-group">
    <h3>Context — OpenAI / Microsoft Tension</h3>
    <ul class="source-list">
      <li>
        <a href="https://www.axios.com/2026/04/13/openai-microsoft-anthropic-amazon" target="_blank" rel="noopener">OpenAI rips Anthropic, distances itself from Microsoft</a>
        <span class="src-meta">Axios &nbsp;·&nbsp; April 13, 2026</span>
      </li>
      <li>
        <a href="https://www.cnbc.com/2026/04/13/openai-touts-amazon-alliance-in-memo-microsoft-limited-our-ability.html" target="_blank" rel="noopener">OpenAI touts Amazon alliance in memo, says Microsoft has "limited our ability" to reach clients</a>
        <span class="src-meta">CNBC &nbsp;·&nbsp; April 13, 2026</span>
      </li>
    </ul>
  </div>

  <div class="source-group">
    <h3>Social</h3>
    <ul class="source-list">
      <li>
        <a href="https://news.ycombinator.com/item?id=47848276" target="_blank" rel="noopener">Anthropic takes $5B from Amazon and pledges $100B in cloud spending in return — discussion</a>
        <span class="src-meta">Hacker News &nbsp;·&nbsp; April 20–22, 2026</span>
      </li>
    </ul>
  </div>