VENTURE CAPITAL | KOCH COMPANY

top of page

“Twin Towers…”

- Vanity Fair

“Twin Towers…”

- Vanity Fair

“Twin Towers…”

- Vanity Fair

“Twin Towers…”

- Vanity Fair

KOCH VENTURE CAPITAL

ALPHABET.

Google’s parent is selling $80B in stock to fund its AI buildout — its first equity raise since 2005, anchored by a $10B private placement from Warren Buffett’s Berkshire Hathaway.

The structure: $10B Berkshire anchor + $30B underwritten (common and mandatory convertible preferred) + $40B sold at-market from Q3. The point of it all: compute. 2026 capex guidance sits at $180–190B, roughly double last year, while Google Cloud carries a $460B backlog on $20B of Q1 revenue (+63%). Demand is outrunning supply.

The product side, from I/O 2026: the Gemini 3.5 series — 3.5 Flash already beats the prior 3.1 Pro on coding and multimodal at 4× the speed, with 3.5 Pro due next month — plus Gemini Omni, the Spark personal agent, Antigravity 2.0, and Android XR glasses this fall.

The science side, from DeepMind: Demis Hassabis’s Nobel-winning AlphaFold now serves 3M+ researchers; Isomorphic Labs targets first human trials by year-end; and a Gemini-integrated autonomous lab is running materials science on its own.

The read: asked what keeps Google up at night, Pichai didn’t say demand — he said compute capacity. The $80B raise is Google answering its own question. It can sell the future faster than it can build the machines to run it.

Sources · Alphabet SEC filing · CNBC · Yahoo Finance · Google I/O 2026 · Fortune · Semafor

—Fromtier-first.

#Alphabet #Google #Gemini #DeepMind #AI

ALPHABET.

Google’s parent is selling $80B in stock to fund its AI buildout — its first equity raise since 2005, anchored by a $10B private placement from Warren Buffett’s Berkshire Hathaway.

The structure: $10B Berkshire anchor + $30B underwritten (common and mandatory convertible preferred) + $40B sold at-market from Q3. The point of it all: compute. 2026 capex guidance sits at $180–190B, roughly double last year, while Google Cloud carries a $460B backlog on $20B of Q1 revenue (+63%). Demand is outrunning supply.

The product side, from I/O 2026: the Gemini 3.5 series — 3.5 Flash already beats the prior 3.1 Pro on coding and multimodal at 4× the speed, with 3.5 Pro due next month — plus Gemini Omni, the Spark personal agent, Antigravity 2.0, and Android XR glasses this fall.

The science side, from DeepMind: Demis Hassabis’s Nobel-winning AlphaFold now serves 3M+ researchers; Isomorphic Labs targets first human trials by year-end; and a Gemini-integrated autonomous lab is running materials science on its own.

The read: asked what keeps Google up at night, Pichai didn’t say demand — he said compute capacity. The $80B raise is Google answering its own question. It can sell the future faster than it can build the machines to run it.

Sources · Alphabet SEC filing · CNBC · Yahoo Finance · Google I/O 2026 · Fortune · Semafor

—Fromtier-first.

#Alphabet #Google #Gemini #DeepMind #AI

SPARK.

Nvidia just declared war on the x86 PC.

At his Computex 2026 keynote, Jensen Huang unveiled the RTX Spark Superchip — a Blackwell GPU and an Arm CPU fused onto one chip, built for laptops and compact desktops.

— Up to 6,144 Blackwell CUDA cores + a 20-core Arm CPU, linked over NVLink-C2C

— 128GB of unified LPDDR5X memory at up to 300 GB/s

— enough to run big models locally

— A Windows-on-Arm platform Nvidia wants to turn into an agentic AI OS

— 30 years of CUDA, RTX, DLSS, FP4, TensorRT and G-SYNC, bundled in

It’s a full-stack move: Nvidia already owns the data center — now it’s coming for the endpoint. Microsoft’s Surface Laptop Ultra is the flagship, and Intel and AMD are suddenly flanked inside the laptop itself.

The roadmap runs three generations deep — Blackwell, then Rubin, then Feynman — with the first Spark systems shipping this fall.

Intel’s reaction? “A healthy dose of paranoia.”

Sources in-frame · full archive in bio.

— Frontier-first.

#Nvidia #RTXSpark #Computex2026 #AI #VC

SPARK.

Nvidia just declared war on the x86 PC.

At his Computex 2026 keynote, Jensen Huang unveiled the RTX Spark Superchip — a Blackwell GPU and an Arm CPU fused onto one chip, built for laptops and compact desktops.

— Up to 6,144 Blackwell CUDA cores + a 20-core Arm CPU, linked over NVLink-C2C

— 128GB of unified LPDDR5X memory at up to 300 GB/s

— enough to run big models locally

— A Windows-on-Arm platform Nvidia wants to turn into an agentic AI OS

— 30 years of CUDA, RTX, DLSS, FP4, TensorRT and G-SYNC, bundled in

It’s a full-stack move: Nvidia already owns the data center — now it’s coming for the endpoint. Microsoft’s Surface Laptop Ultra is the flagship, and Intel and AMD are suddenly flanked inside the laptop itself.

The roadmap runs three generations deep — Blackwell, then Rubin, then Feynman — with the first Spark systems shipping this fall.

Intel’s reaction? “A healthy dose of paranoia.”

Sources in-frame · full archive in bio.

— Frontier-first.

#Nvidia #RTXSpark #Computex2026 #AI #VC

DELL.

Best day ever. The “old PC company” just got repriced as the backbone of the AI buildout.

— Q1 FY27 revenue hit $43.8B, up 88% — blowing past the ~$35.8B Street estimate.

— Adjusted EPS of $4.86 vs $2.94 expected — up 214% year-over-year.

— AI server revenue exploded 757% to $16.1B.

— $24.4B in new AI orders and a record $51.3B backlog

— with a pipeline still multiple times larger.

— FY27 guidance raised to $165–169B, with ~$60B in AI servers.

The result: +32% in a single session — the biggest day in company history — and a stock up 234% on the year, touching an all-time high near $467.

The catch: it’s parabolic. ~38× forward earnings, +330% off the lows, and Super Micro right behind it. Vertical moves cut both ways.

“Never seen anything like it.” — Ben Reitzes, Melius

Sources in-frame · full archive in bio.

Personal commentary. Not investment advice. Not an offer or solicitation.

— Frontier-first.

#DELL #AI #AIinfrastructure #datacenter #stockmarke

DELL.

Best day ever. The “old PC company” just got repriced as the backbone of the AI buildout.

— Q1 FY27 revenue hit $43.8B, up 88% — blowing past the ~$35.8B Street estimate.

— Adjusted EPS of $4.86 vs $2.94 expected — up 214% year-over-year.

— AI server revenue exploded 757% to $16.1B.

— $24.4B in new AI orders and a record $51.3B backlog

— with a pipeline still multiple times larger.

— FY27 guidance raised to $165–169B, with ~$60B in AI servers.

The result: +32% in a single session — the biggest day in company history — and a stock up 234% on the year, touching an all-time high near $467.

The catch: it’s parabolic. ~38× forward earnings, +330% off the lows, and Super Micro right behind it. Vertical moves cut both ways.

“Never seen anything like it.” — Ben Reitzes, Melius

Sources in-frame · full archive in bio.

Personal commentary. Not investment advice. Not an offer or solicitation.

— Frontier-first.

#DELL #AI #AIinfrastructure #datacenter #stockmarke

IPO.

The Claude maker confidentially filed a draft S-1 with the SEC today — the first formal step toward what could be one of the largest tech listings ever.

Confidential means no public numbers yet: share count and price aren’t set, and the filing simply gives Anthropic the option to go public once the SEC completes its review. It’s optionality, not a commitment — and widely seen as its last move as a private company.

The backdrop: a $65B Series H closed May 28 at a $965B post-money valuation, edging past OpenAI’s $852B for the first time. Run-rate revenue crossed $47B in May, up from ~$14B in February — a 5.3× valuation climb in roughly nine months.

But a confidential S-1 starts the clock; it doesn’t stop it. The audited revenue, gross margins and cash burn that public investors actually price won’t surface until the filing goes public. Until then, $965B is a headline, not yet an investment case.

Anthropic now joins SpaceX (debut June 12) and a still-private OpenAI in the most dramatic frontier-AI IPO wave on record.

Sources · Anthropic (Rule 135) · CNBC · TechCrunch · FT · Bloomberg

Personal commentary. Not investment advice. Not an offer or solicitation.

— Frontier-first.

#Anthropic #Claude #IPO #AI #VentureCapital

IPO.

The Claude maker confidentially filed a draft S-1 with the SEC today — the first formal step toward what could be one of the largest tech listings ever.

Confidential means no public numbers yet: share count and price aren’t set, and the filing simply gives Anthropic the option to go public once the SEC completes its review. It’s optionality, not a commitment — and widely seen as its last move as a private company.

The backdrop: a $65B Series H closed May 28 at a $965B post-money valuation, edging past OpenAI’s $852B for the first time. Run-rate revenue crossed $47B in May, up from ~$14B in February — a 5.3× valuation climb in roughly nine months.

But a confidential S-1 starts the clock; it doesn’t stop it. The audited revenue, gross margins and cash burn that public investors actually price won’t surface until the filing goes public. Until then, $965B is a headline, not yet an investment case.

Anthropic now joins SpaceX (debut June 12) and a still-private OpenAI in the most dramatic frontier-AI IPO wave on record.

Sources · Anthropic (Rule 135) · CNBC · TechCrunch · FT · Bloomberg

Personal commentary. Not investment advice. Not an offer or solicitation.

— Frontier-first.

#Anthropic #Claude #IPO #AI #VentureCapital

MAY 2026.

A month to remember.

— SPACEXAI files its S-1: the largest IPO in history. $1.75T target, first trade June 12.

— CEREBRAS debuts on Nasdaq and pops 68% on Day One — the biggest US tech IPO since Uber.

— ANTHROPIC closes $65B at a $965B valuation, eclipsing OpenAI for the first time.

— NVIDIA prints $81.6B in a single quarter and guides to $91B for the next.

— ANDURIL doubles to $61B — the largest defense-tech raise on record.

— BLUE ORIGIN’s New Glenn explodes on its only orbital pad; the space basket tumbles.

Swipe to see who else is making airwaves — chips, defense, quantum, rare earths, prediction markets, and the rest of the map that moved this month. 👉

Sources in-frame · full archive in bio.

Personal commentary. Not investment advice. Not an offer or solicitation.

— Frontier-first.

#AI #frontiertech #spacetech #defensetech #venturecapital

MAY 2026.

A month to remember.

— SPACEXAI files its S-1: the largest IPO in history. $1.75T target, first trade June 12.

— CEREBRAS debuts on Nasdaq and pops 68% on Day One — the biggest US tech IPO since Uber.

— ANTHROPIC closes $65B at a $965B valuation, eclipsing OpenAI for the first time.

— NVIDIA prints $81.6B in a single quarter and guides to $91B for the next.

— ANDURIL doubles to $61B — the largest defense-tech raise on record.

— BLUE ORIGIN’s New Glenn explodes on its only orbital pad; the space basket tumbles.

Swipe to see who else is making airwaves — chips, defense, quantum, rare earths, prediction markets, and the rest of the map that moved this month. 👉

Sources in-frame · full archive in bio.

Personal commentary. Not investment advice. Not an offer or solicitation.

— Frontier-first.

#AI #frontiertech #spacetech #defensetech #venturecapital

ANTHROPIC.

Anthropic just closed a $65B Series H at a $965B post-money valuation — eclipsing OpenAI for the first time. One of the largest private raises in tech history, nearly tripling the company’s value in 90 days.

The round (May 28):

· Co-led by Altimeter, Dragoneer, Greenoaks, and Sequoia — each putting in >$2B (Sequoia’s biggest check ever)
· Institutional roster: Baillie Gifford, Blackstone, Brookfield, D.E. Shaw, DST Global, Fidelity
· Strategic hardware partners: Micron, Samsung, SK Hynix — the HBM stack AI training depends on
· ~$15B is previously committed hyperscaler money, including $5B from Amazon

The trajectory:

· Feb 2026: ~$380B → May 2026: $965B (nearly 3× in 90 days)
· Run-rate revenue crossed $47B earlier this month
· A projected 130% revenue surge would deliver Anthropic’s first operating profit

The overtake:

· Anthropic $965B post-money vs OpenAI $852B (after OpenAI’s record $122B March raise)
· First time Anthropic’s private valuation sits above OpenAI’s
· Honest balance: OpenAI’s consumer reach (ChatGPT) is far larger; operating businesses still distinct shapes
The IPO question:
· TechCrunch framed this as “possibly the last private round before an IPO” — outlet framing, not company guidance
· Wilson Sonsini engaged on early IPO prep (Jan 2026); cash-flow-positive target pushed to 2028
· At ~$47B run-rate and ~$965B, public markets become a real option — not the only one

The honest risk: ~20× sales is rich; a reported $1.25B/month SpaceX compute deal underscores enormous capital intensity. Execution matters now.

Brad Gerstner (Altimeter): “Claude’s adoption among the most demanding enterprise customers positions Anthropic to lead the next phase of AI innovation.”

— Frontier-first.

Sources: Reuters · Bloomberg · TechCrunch · CNBC · Axios · AI Insider

#Anthropic #Claude #AI #SeriesH #Frontier

ANTHROPIC.

Anthropic just closed a $65B Series H at a $965B post-money valuation — eclipsing OpenAI for the first time. One of the largest private raises in tech history, nearly tripling the company’s value in 90 days.

The round (May 28):

· Co-led by Altimeter, Dragoneer, Greenoaks, and Sequoia — each putting in >$2B (Sequoia’s biggest check ever)
· Institutional roster: Baillie Gifford, Blackstone, Brookfield, D.E. Shaw, DST Global, Fidelity
· Strategic hardware partners: Micron, Samsung, SK Hynix — the HBM stack AI training depends on
· ~$15B is previously committed hyperscaler money, including $5B from Amazon

The trajectory:

· Feb 2026: ~$380B → May 2026: $965B (nearly 3× in 90 days)
· Run-rate revenue crossed $47B earlier this month
· A projected 130% revenue surge would deliver Anthropic’s first operating profit

The overtake:

· Anthropic $965B post-money vs OpenAI $852B (after OpenAI’s record $122B March raise)
· First time Anthropic’s private valuation sits above OpenAI’s
· Honest balance: OpenAI’s consumer reach (ChatGPT) is far larger; operating businesses still distinct shapes
The IPO question:
· TechCrunch framed this as “possibly the last private round before an IPO” — outlet framing, not company guidance
· Wilson Sonsini engaged on early IPO prep (Jan 2026); cash-flow-positive target pushed to 2028
· At ~$47B run-rate and ~$965B, public markets become a real option — not the only one

The honest risk: ~20× sales is rich; a reported $1.25B/month SpaceX compute deal underscores enormous capital intensity. Execution matters now.

Brad Gerstner (Altimeter): “Claude’s adoption among the most demanding enterprise customers positions Anthropic to lead the next phase of AI innovation.”

— Frontier-first.

Sources: Reuters · Bloomberg · TechCrunch · CNBC · Axios · AI Insider

#Anthropic #Claude #AI #SeriesH #Frontier

BLUE ORIGIN.

Bezos’s New Glenn exploded on its only orbital pad during a hotfire test last night. No injuries — but space stocks tumbled today and the fallout reaches Artemis, Amazon, and the whole orbit economy.

The explosion (9 PM EDT, May 28):

· Static fire of New Glenn’s seven BE-4 first-stage engines at Cape Canaveral LC-36
· As the engines fired, the 188-ft first stage was engulfed in flame; methane + LOX ignited in a fireball
· Intended cargo: Amazon “Project Kuiper” satellites — the Starlink challenger
Why this hurts more than a normal loss:
· LC-36 is Blue Origin’s ONLY orbital pad — the program is grounded until it’s rebuilt
· Second New Glenn anomaly in two months (April’s flight had a 2nd-stage failure)
· SpaceX has multiple pads (Vandenberg, KSC 39A, Starbase); Blue Origin has one

The market verdict (Friday):

· 📉 AST SpaceMobile (ASTS) ~-17%; Deutsche Bank cut to Hold, PT $117→$106 (sees ~6-mo slip)
· Redwire -14% · Planet Labs -8% · Rocket Lab -6% · Intuitive Machines -5% · Voyager -6%
· Procure Space ETF (UFO) on pace for its worst session of 2026
· The twist: Virgin Galactic +11% (positioning for a New Shepard tourism shift)

The bigger stakes:

· Artemis — Blue Moon was the official backup lunar lander; 2028 NASA option narrows
· Kuiper has FCC deadlines; Amazon may have to lean on Vulcan or even SpaceX
· The ORBIT ECONOMY thesis modeled on TWO heavy-lift providers is more fragile today

Bezos: “Very rough day, but we’ll rebuild whatever needs rebuilding and get back to flying. It’s worth it.”

— Frontier-first.

Sources: Spaceflight Now · Yahoo Finance · Seeking Alpha · TipRanks · 24/7 Wall St

#BlueOrigin #NewGlenn #SpaceX #ASTS #Artemis

BLUE ORIGIN.

Bezos’s New Glenn exploded on its only orbital pad during a hotfire test last night. No injuries — but space stocks tumbled today and the fallout reaches Artemis, Amazon, and the whole orbit economy.

The explosion (9 PM EDT, May 28):

· Static fire of New Glenn’s seven BE-4 first-stage engines at Cape Canaveral LC-36
· As the engines fired, the 188-ft first stage was engulfed in flame; methane + LOX ignited in a fireball
· Intended cargo: Amazon “Project Kuiper” satellites — the Starlink challenger
Why this hurts more than a normal loss:
· LC-36 is Blue Origin’s ONLY orbital pad — the program is grounded until it’s rebuilt
· Second New Glenn anomaly in two months (April’s flight had a 2nd-stage failure)
· SpaceX has multiple pads (Vandenberg, KSC 39A, Starbase); Blue Origin has one

The market verdict (Friday):

· 📉 AST SpaceMobile (ASTS) ~-17%; Deutsche Bank cut to Hold, PT $117→$106 (sees ~6-mo slip)
· Redwire -14% · Planet Labs -8% · Rocket Lab -6% · Intuitive Machines -5% · Voyager -6%
· Procure Space ETF (UFO) on pace for its worst session of 2026
· The twist: Virgin Galactic +11% (positioning for a New Shepard tourism shift)

The bigger stakes:

· Artemis — Blue Moon was the official backup lunar lander; 2028 NASA option narrows
· Kuiper has FCC deadlines; Amazon may have to lean on Vulcan or even SpaceX
· The ORBIT ECONOMY thesis modeled on TWO heavy-lift providers is more fragile today

Bezos: “Very rough day, but we’ll rebuild whatever needs rebuilding and get back to flying. It’s worth it.”

— Frontier-first.

Sources: Spaceflight Now · Yahoo Finance · Seeking Alpha · TipRanks · 24/7 Wall St

#BlueOrigin #NewGlenn #SpaceX #ASTS #Artemis

SAAS.

Agentic AI is breaking the per-seat model that built a trillion-dollar industry. Software is being repriced from access to output — in real time.

The break:

· ~$285B erased from SaaS stocks in one Feb wave; ~$2T in broader software repricing by mid-March 2026
· Jason Lemkin: “If 10 agents do the work of 100 reps, you need 10 seats — not 100.”
· Monday.com replaced 100 SDRs with agents. Atlassian’s first-ever seat decline. Klarna cut ~700.
· Gartner: 40% of SaaS spend shifts to usage/outcome by 2030. Bloomberg: subscriptions fall 60% → 30%.

The new pricing ladder:

· Consumption — per action/token (Copilot Studio $0.01/credit; Salesforce $0.10/action)
· Outcome — per result (Intercom Fin $0.99/resolution; Zendesk $1.50–2.00)
· Hybrid — ~65% of vendors bolted an “AI meter” onto seats. The “AI Tax.”
Salesforce — the boldest pivot. The most seat-dependent giant moved seat → outcome with Agentforce (3.2 TRILLION tokens in Q3 2026) plus an Agentic Enterprise License Agreement for unlimited use.

Snowflake — the quiet winner. Consumption-native for years. Agents are heavy compute/data consumers — exactly what Snowflake meters.

The cost trap (consumption cuts both ways):

· Uber burned its ENTIRE 2026 AI coding budget in 4 months — $500–$2,000/engineer/mo; 95% use AI tools monthly; ~11% of backend code now AI-written
· Uber’s COO on ROI: “it’s very hard to draw a line” — back to the drawing board on budgeting
· Microsoft pulled internal Claude Code licenses in one division, steering staff to GitHub Copilot. Same tool, opposite call.

Winners price for output. Pure per-seat vendors bleed. The clearest winners may sit underneath — the model labs and infra metering the tokens.

When the work is done by software, you’re not charging for access anymore. You’re charging for outcomes.

— Frontier-first.

Personal commentary. Not investment advice. Not an offer or solicitation.

Sources: SaaStr · Bain · Gartner · Fortune · The SaaS CFO · 05.28.26

#SaaS #AgenticAI #Salesforce #Snowflake #Frontier

SAAS.

Agentic AI is breaking the per-seat model that built a trillion-dollar industry. Software is being repriced from access to output — in real time.

The break:

· ~$285B erased from SaaS stocks in one Feb wave; ~$2T in broader software repricing by mid-March 2026
· Jason Lemkin: “If 10 agents do the work of 100 reps, you need 10 seats — not 100.”
· Monday.com replaced 100 SDRs with agents. Atlassian’s first-ever seat decline. Klarna cut ~700.
· Gartner: 40% of SaaS spend shifts to usage/outcome by 2030. Bloomberg: subscriptions fall 60% → 30%.

The new pricing ladder:

· Consumption — per action/token (Copilot Studio $0.01/credit; Salesforce $0.10/action)
· Outcome — per result (Intercom Fin $0.99/resolution; Zendesk $1.50–2.00)
· Hybrid — ~65% of vendors bolted an “AI meter” onto seats. The “AI Tax.”
Salesforce — the boldest pivot. The most seat-dependent giant moved seat → outcome with Agentforce (3.2 TRILLION tokens in Q3 2026) plus an Agentic Enterprise License Agreement for unlimited use.

Snowflake — the quiet winner. Consumption-native for years. Agents are heavy compute/data consumers — exactly what Snowflake meters.

The cost trap (consumption cuts both ways):

· Uber burned its ENTIRE 2026 AI coding budget in 4 months — $500–$2,000/engineer/mo; 95% use AI tools monthly; ~11% of backend code now AI-written
· Uber’s COO on ROI: “it’s very hard to draw a line” — back to the drawing board on budgeting
· Microsoft pulled internal Claude Code licenses in one division, steering staff to GitHub Copilot. Same tool, opposite call.

Winners price for output. Pure per-seat vendors bleed. The clearest winners may sit underneath — the model labs and infra metering the tokens.

When the work is done by software, you’re not charging for access anymore. You’re charging for outcomes.

— Frontier-first.

Personal commentary. Not investment advice. Not an offer or solicitation.

Sources: SaaStr · Bain · Gartner · Fortune · The SaaS CFO · 05.28.26

#SaaS #AgenticAI #Salesforce #Snowflake #Frontier

STARGATE.

OpenAI’s $500B megaproject hit 7 gigawatts ahead of schedule. But the financial model is wobbling. The most ambitious infrastructure bet in tech history — and its first cracks.

The scale (ahead of schedule):

· Original commitment: $500B for 10 GW over four years (announced at the White House, Jan 2025)
· Now: nearly 7 GW planned or under construction, $400B+ over three years
· GPT-5.5 was trained at the Abilene flagship — the compute is real
· Backers: OpenAI leads, Oracle builds, SoftBank finances

Abilene, TX — the flagship:

· 450,000+ Nvidia GB200 GPUs under a 15-year Oracle lease
· Ten half-million-sq-ft buildings on 1,000+ acres; ~1 GW by mid-2026
· Operated by Crusoe Energy (flare-gas-to-AI pivot); ~$15B total — the most expensive data center ever financed

The national map:

· Oracle: Shackelford TX, Doña Ana NM, Wisconsin, + Michigan, Wyoming, Pennsylvania
· SoftBank + OpenAI: Lordstown OH, Milam County TX (scaling to 1.5 GW)
· 25,000+ onsite jobs projected

The first cracks:

· March 2026: Oracle + OpenAI abandoned the Abilene 1.2→2.0 GW expansion over financing terms and “often-changing demand forecasting” (Bloomberg)
· The joint venture stalled over control disputes — reportedly no staff hired, no centers built under the JV
· OpenAI reframed spend from $1.4T to $600B and is “now renting what it vowed to build” (AWS, Google Cloud, CoreWeave)

The big question:

· $1.4T+ in nameplate commitments vs ~$20B 2026 revenue
· Circular-financing critique: Nvidia invests in OpenAI, which buys Nvidia chips, in centers Nvidia also funds
· Altman & Friar called a WSJ report “ridiculous,” citing $122B in cumulative funding

The catalyst:

· Anticipated Q4 2026 IPO targeting a $1 trillion valuation
· $280B projected annual revenue by 2030
· Layered $300B, five-year Oracle cloud commitment (2027–2032)

— Frontier-first.

Sources: OpenAI · Bloomberg · CNBC · DCD · Reuters · IntuitionLabs · Tech-Insider · TechTimes · 05.21.26

#Stargate #OpenAI #AIInfrastructure #1789Capital #Frontier

STARGATE.

OpenAI’s $500B megaproject hit 7 gigawatts ahead of schedule. But the financial model is wobbling. The most ambitious infrastructure bet in tech history — and its first cracks.

The scale (ahead of schedule):

· Original commitment: $500B for 10 GW over four years (announced at the White House, Jan 2025)
· Now: nearly 7 GW planned or under construction, $400B+ over three years
· GPT-5.5 was trained at the Abilene flagship — the compute is real
· Backers: OpenAI leads, Oracle builds, SoftBank finances

Abilene, TX — the flagship:

· 450,000+ Nvidia GB200 GPUs under a 15-year Oracle lease
· Ten half-million-sq-ft buildings on 1,000+ acres; ~1 GW by mid-2026
· Operated by Crusoe Energy (flare-gas-to-AI pivot); ~$15B total — the most expensive data center ever financed

The national map:

· Oracle: Shackelford TX, Doña Ana NM, Wisconsin, + Michigan, Wyoming, Pennsylvania
· SoftBank + OpenAI: Lordstown OH, Milam County TX (scaling to 1.5 GW)
· 25,000+ onsite jobs projected

The first cracks:

· March 2026: Oracle + OpenAI abandoned the Abilene 1.2→2.0 GW expansion over financing terms and “often-changing demand forecasting” (Bloomberg)
· The joint venture stalled over control disputes — reportedly no staff hired, no centers built under the JV
· OpenAI reframed spend from $1.4T to $600B and is “now renting what it vowed to build” (AWS, Google Cloud, CoreWeave)

The big question:

· $1.4T+ in nameplate commitments vs ~$20B 2026 revenue
· Circular-financing critique: Nvidia invests in OpenAI, which buys Nvidia chips, in centers Nvidia also funds
· Altman & Friar called a WSJ report “ridiculous,” citing $122B in cumulative funding

The catalyst:

· Anticipated Q4 2026 IPO targeting a $1 trillion valuation
· $280B projected annual revenue by 2030
· Layered $300B, five-year Oracle cloud commitment (2027–2032)

— Frontier-first.

Sources: OpenAI · Bloomberg · CNBC · DCD · Reuters · IntuitionLabs · Tech-Insider · TechTimes · 05.21.26

#Stargate #OpenAI #AIInfrastructure #1789Capital #Frontier

ROBINHOOD.

Retail just got what Wall Street has had for years. Robinhood opened its platform to autonomous AI agents today — the line from “AI analyzes my money” to “AI moves my money” just got crossed.

The line that just got crossed:

· Until today, every AI-finance integration was read-only. ChatGPT + Plaid: analyze your accounts. Perplexity Computer + Plaid: dashboards & budgets. Ramp: agents that automate corporate finance ops.
· Robinhood Agentic Trading authorizes AI to ANALYZE portfolios, REBALANCE holdings, DEPLOY strategies, and EXECUTE trades — autonomously.
· The PLAID Trust Gap thesis (55% use AI for finance, 18% trust it for big decisions) just narrowed by months in a single announcement.

How it works:

· Customers open a dedicated agentic account — isolated from their main portfolio. The agent can only touch funds the user deposits.
· Built on Model Context Protocol (MCP) — the open standard introduced by Anthropic. Users can plug in any agent from any platform.
· Equities first, in beta. Options, crypto, futures, event contracts, and prediction markets to follow.
· Push notifications on every trade, optional manual previews, instant kill switch.

The wallet twin — Agentic Credit Card:

· Separate virtual cards your AI agent can spend from, with custom caps and optional approval per purchase.
· 3% cash back on Robinhood Gold for agentic purchases.
· The consumer mirror of Ramp’s corporate playbook — virtual cards, granular controls, autonomous spend. Ramp proved the model for finance teams; Robinhood is scaling it to 27M retail customers.

The honest concern:

· Putting autonomous trading in retail hands — without the institutional risk controls hedge funds use — raises real questions. Hallucinated trades, prompt injection, agent loops. New surface area, new failure modes.
· Robinhood’s guardrails: fraud detection, manual previews, spending caps, push notifications, one-tap shutdown.

— Frontier-first.

Sources: BB· CNBC · TechCrunch · Axios · Fortune · Robinhood Statement · Crypto Briefing

#Robinhood #HOOD #AgenticFinance #Plaid #Ramp

ROBINHOOD.

Retail just got what Wall Street has had for years. Robinhood opened its platform to autonomous AI agents today — the line from “AI analyzes my money” to “AI moves my money” just got crossed.

The line that just got crossed:

· Until today, every AI-finance integration was read-only. ChatGPT + Plaid: analyze your accounts. Perplexity Computer + Plaid: dashboards & budgets. Ramp: agents that automate corporate finance ops.
· Robinhood Agentic Trading authorizes AI to ANALYZE portfolios, REBALANCE holdings, DEPLOY strategies, and EXECUTE trades — autonomously.
· The PLAID Trust Gap thesis (55% use AI for finance, 18% trust it for big decisions) just narrowed by months in a single announcement.

How it works:

· Customers open a dedicated agentic account — isolated from their main portfolio. The agent can only touch funds the user deposits.
· Built on Model Context Protocol (MCP) — the open standard introduced by Anthropic. Users can plug in any agent from any platform.
· Equities first, in beta. Options, crypto, futures, event contracts, and prediction markets to follow.
· Push notifications on every trade, optional manual previews, instant kill switch.

The wallet twin — Agentic Credit Card:

· Separate virtual cards your AI agent can spend from, with custom caps and optional approval per purchase.
· 3% cash back on Robinhood Gold for agentic purchases.
· The consumer mirror of Ramp’s corporate playbook — virtual cards, granular controls, autonomous spend. Ramp proved the model for finance teams; Robinhood is scaling it to 27M retail customers.

The honest concern:

· Putting autonomous trading in retail hands — without the institutional risk controls hedge funds use — raises real questions. Hallucinated trades, prompt injection, agent loops. New surface area, new failure modes.
· Robinhood’s guardrails: fraud detection, manual previews, spending caps, push notifications, one-tap shutdown.

— Frontier-first.

Sources: BB· CNBC · TechCrunch · Axios · Fortune · Robinhood Statement · Crypto Briefing

#Robinhood #HOOD #AgenticFinance #Plaid #Ramp

RAMP.

Ramp is in talks to raise $750M at a $40B+ valuation — its fifth step-up in 14 months. The AI-first fintech is pulling away from the entire pack.

The run (the valuation rocket):

· March 2025: $13B → June: $16B (Founders Fund) → July: $22.5B (Iconiq) → November: $32B (Lightspeed) → now: $40B+ in talks
· Five valuation step-ups in 14 months — more than tripling its worth
· The new ~$750M round is reportedly co-led by Iconiq and Singapore’s sovereign fund GIC
· At $40B+, within shouting distance of Stripe in the private fintech league tables

The platform:

· 50,000+ companies; crossed $1B in annualized recurring revenue (Sept 2025)
· Started in 2019 as a corporate card — now a full finance-ops hub: cards, expense tracking, accounting, AP, procurement, treasury
· Founders Eric Glyman & Karim Atiyeh (who sold Paribus to Capital One); the thesis from day one: software that saves companies money

The AI edge (where it actually lands):

· Autonomous agents now handle real-time fraud prevention, policy enforcement, and cash-flow optimization
· Receipts reviewed, expenses categorized, accounts reconciled — automatically
· The investor lens has shifted: capital now asks whether AI truly replaces labor inside the workflow — and whether it shows up in unit economics. Ramp’s do.

The knockout:

· Brex — Ramp’s longtime rival — faced layoffs and was acquired by Capital One for ~$5.15B in 2026
· Airbase, BILL, Rho, SAP Concur offer fragmented tools; Ramp’s edge is end-to-end automation across the whole spend lifecycle
· As peers consolidate or stall, Ramp more than tripled its valuation

The read: whoever owns the daily finance workflow increasingly owns the funding conversation that comes with it.

— Frontier-first.

Sources: TechCrunch · WSJ · CNBC Disruptor 50 · Funder Intel · Crowdfund Insider · FinTech Futures · 05.26.26

#Ramp #Fintech #AI #VentureCapital #aistartup

RAMP.

Ramp is in talks to raise $750M at a $40B+ valuation — its fifth step-up in 14 months. The AI-first fintech is pulling away from the entire pack.

The run (the valuation rocket):

· March 2025: $13B → June: $16B (Founders Fund) → July: $22.5B (Iconiq) → November: $32B (Lightspeed) → now: $40B+ in talks
· Five valuation step-ups in 14 months — more than tripling its worth
· The new ~$750M round is reportedly co-led by Iconiq and Singapore’s sovereign fund GIC
· At $40B+, within shouting distance of Stripe in the private fintech league tables

The platform:

· 50,000+ companies; crossed $1B in annualized recurring revenue (Sept 2025)
· Started in 2019 as a corporate card — now a full finance-ops hub: cards, expense tracking, accounting, AP, procurement, treasury
· Founders Eric Glyman & Karim Atiyeh (who sold Paribus to Capital One); the thesis from day one: software that saves companies money

The AI edge (where it actually lands):

· Autonomous agents now handle real-time fraud prevention, policy enforcement, and cash-flow optimization
· Receipts reviewed, expenses categorized, accounts reconciled — automatically
· The investor lens has shifted: capital now asks whether AI truly replaces labor inside the workflow — and whether it shows up in unit economics. Ramp’s do.

The knockout:

· Brex — Ramp’s longtime rival — faced layoffs and was acquired by Capital One for ~$5.15B in 2026
· Airbase, BILL, Rho, SAP Concur offer fragmented tools; Ramp’s edge is end-to-end automation across the whole spend lifecycle
· As peers consolidate or stall, Ramp more than tripled its valuation

The read: whoever owns the daily finance workflow increasingly owns the funding conversation that comes with it.

— Frontier-first.

Sources: TechCrunch · WSJ · CNBC Disruptor 50 · Funder Intel · Crowdfund Insider · FinTech Futures · 05.26.26

#Ramp #Fintech #AI #VentureCapital #aistartup

MICRON.

Micron just became the first memory chipmaker in history to cross $1 trillion. The stock popped 18% today as the AI trade broadens past the GPU.

The milestone (today, May 26, 2026):

· MU jumped 18% to ~$886/share, pushing market cap past $1 trillion for the first time
· The first pure-play memory company ever to reach the milestone
· Up ~70% in 2026 alone — outpacing the broader chip indexes and every rival memory maker
· Market cap has multiplied several-fold from the ~$100B range a couple years ago

The catalyst:

· UBS sharply raised its target, forecasting ~90% upside over the next year — citing stronger AI demand and long-term supply deals
· The revised target is the highest of the 46 brokerages covering MU — implying a valuation near $1.8 trillion within 12 months (up from $847B last Friday)
· A political tailwind too: Trump gave Micron a shoutout at a rally Friday

Why memory (the real story):

· For two years, Nvidia WAS the AI trade. Now investors realize compute is useless without memory to feed it
· Micron’s high-bandwidth memory (HBM) — the stacked memory beside AI GPUs — is sold out
· When availability, not price, is the constraint, the supplier gets pricing power
· Micron announced a ~$200B capex plan to expand memory fabrication

The fundamentals (not hype):

· Q1 FY2026 EPS: $12.20 vs $9.19 expected — a 33% earnings surprise
· Gross margin expanded to 46.7%, up from 35% a year prior — even while pouring cash into fabs
· The honest caveat: memory is historically cyclical; the bull case rests on AI breaking the old boom-bust pattern

The read:

“In the AI era, memory is no longer a commodity. It’s the bottleneck.”

— Frontier-first.

Personal commentary. Not investment advice. Not an offer or solicitation.

Sources: CNBC · Reuters · UBS · Investing.com · GuruFocus · Motley Fool · Yahoo Finance · 05.26.26

#Micron #MU #Semiconductors #AI #HBM

MICRON.

Micron just became the first memory chipmaker in history to cross $1 trillion. The stock popped 18% today as the AI trade broadens past the GPU.

The milestone (today, May 26, 2026):

· MU jumped 18% to ~$886/share, pushing market cap past $1 trillion for the first time
· The first pure-play memory company ever to reach the milestone
· Up ~70% in 2026 alone — outpacing the broader chip indexes and every rival memory maker
· Market cap has multiplied several-fold from the ~$100B range a couple years ago

The catalyst:

· UBS sharply raised its target, forecasting ~90% upside over the next year — citing stronger AI demand and long-term supply deals
· The revised target is the highest of the 46 brokerages covering MU — implying a valuation near $1.8 trillion within 12 months (up from $847B last Friday)
· A political tailwind too: Trump gave Micron a shoutout at a rally Friday

Why memory (the real story):

· For two years, Nvidia WAS the AI trade. Now investors realize compute is useless without memory to feed it
· Micron’s high-bandwidth memory (HBM) — the stacked memory beside AI GPUs — is sold out
· When availability, not price, is the constraint, the supplier gets pricing power
· Micron announced a ~$200B capex plan to expand memory fabrication

The fundamentals (not hype):

· Q1 FY2026 EPS: $12.20 vs $9.19 expected — a 33% earnings surprise
· Gross margin expanded to 46.7%, up from 35% a year prior — even while pouring cash into fabs
· The honest caveat: memory is historically cyclical; the bull case rests on AI breaking the old boom-bust pattern

The read:

“In the AI era, memory is no longer a commodity. It’s the bottleneck.”

— Frontier-first.

Personal commentary. Not investment advice. Not an offer or solicitation.

Sources: CNBC · Reuters · UBS · Investing.com · GuruFocus · Motley Fool · Yahoo Finance · 05.26.26

#Micron #MU #Semiconductors #AI #HBM

ORBIT ECONOMY.

NASA is handing low Earth orbit to private industry. Four startups are building the hardware and habitats of the post-ISS era — and the race to own the orbital economy is on.

The shift:

· NASA is moving from owning a space station to buying services from commercial providers — to cut costs and refocus on Artemis
· The ISS is funded through ~2030; the race is to have private stations flying before then
· It all rides on cheap heavy lift — SpaceX’s Starship and Blue Origin’s New Glenn

APEX (Los Angeles) — the fast builder:

· First manufacturer of productized, high-rate, configurable satellite buses — the off-the-shelf chassis others build on
· $200M Series D (Sept 2025) → $1B+ valuation; a unicorn 3 years to the day after founding
· Defense is ~2/3 of business: $45.9M Space Force contract, plus BAE Systems and Anduril as customers

K2 SPACE (Torrance) — the go-big bet:

· Founded by ex-SpaceX engineers; builds giant “Mega Class” buses (~10× the power of their class)
· GRAVITAS launched March 2026 — 2 tons, 40m solar wingspan, 20 kW; $250M Series C at a $3B valuation
· Next: “Giga” — a 100 kW platform for Starship/New Glenn, aimed squarely at on-orbit compute

AXIOM (Houston) — the habitat path:

· Attaches modules to the ISS first, then separates into a free-flyer as early as 2028
· Most proven operator: 4 private astronaut missions flown, a 5th awarded Jan 2026
· Honest tension: CEO has said the market may not support more than one station; capital has been tight

STARLAB — the single-launch lab:

· A Voyager Technologies + Airbus JV, with Northrop Grumman support
· The audacious bet: deploy the entire station in ONE Starship launch (2028), resupplied by Cygnus; ops ~2029

The stack: hardware below (Apex + K2), habitats above (Axiom + Starlab). And K2’s 100 kW Giga is the bridge to orbital data centers — when power is the constraint, put the compute where uninterrupted solar lives. Sea, land, and now space.

— Frontier-first.

Sources: NASA · SpaceNews · PRNewswire · SatNews · Payload · NSS · Orbital Radar · NASASpaceflight · MLQ.ai · 05.26.26

#Aerospace #ApexSpace #K2Space #AxiomSpace #Starlab

ORBIT ECONOMY.

NASA is handing low Earth orbit to private industry. Four startups are building the hardware and habitats of the post-ISS era — and the race to own the orbital economy is on.

The shift:

· NASA is moving from owning a space station to buying services from commercial providers — to cut costs and refocus on Artemis
· The ISS is funded through ~2030; the race is to have private stations flying before then
· It all rides on cheap heavy lift — SpaceX’s Starship and Blue Origin’s New Glenn

APEX (Los Angeles) — the fast builder:

· First manufacturer of productized, high-rate, configurable satellite buses — the off-the-shelf chassis others build on
· $200M Series D (Sept 2025) → $1B+ valuation; a unicorn 3 years to the day after founding
· Defense is ~2/3 of business: $45.9M Space Force contract, plus BAE Systems and Anduril as customers

K2 SPACE (Torrance) — the go-big bet:

· Founded by ex-SpaceX engineers; builds giant “Mega Class” buses (~10× the power of their class)
· GRAVITAS launched March 2026 — 2 tons, 40m solar wingspan, 20 kW; $250M Series C at a $3B valuation
· Next: “Giga” — a 100 kW platform for Starship/New Glenn, aimed squarely at on-orbit compute

AXIOM (Houston) — the habitat path:

· Attaches modules to the ISS first, then separates into a free-flyer as early as 2028
· Most proven operator: 4 private astronaut missions flown, a 5th awarded Jan 2026
· Honest tension: CEO has said the market may not support more than one station; capital has been tight

STARLAB — the single-launch lab:

· A Voyager Technologies + Airbus JV, with Northrop Grumman support
· The audacious bet: deploy the entire station in ONE Starship launch (2028), resupplied by Cygnus; ops ~2029

The stack: hardware below (Apex + K2), habitats above (Axiom + Starlab). And K2’s 100 kW Giga is the bridge to orbital data centers — when power is the constraint, put the compute where uninterrupted solar lives. Sea, land, and now space.

— Frontier-first.

Sources: NASA · SpaceNews · PRNewswire · SatNews · Payload · NSS · Orbital Radar · NASASpaceflight · MLQ.ai · 05.26.26

#Aerospace #ApexSpace #K2Space #AxiomSpace #Starlab

PANTHALASSA.

Floating AI data centers powered entirely by ocean waves. No grid. No anchors. Peter Thiel just led a $140M round at a reported $1 billion valuation.

The concept:

· Ocean-3 — an autonomous, self-propelled floating platform that generates clean wave energy to power a modular AI data center
· No anchors, no cables to land or seabed — a fully self-contained unit
· Power comes from waves: water forced through an internal turbine as the platform rises and falls, sited in the planet’s most energy-dense wave regions
· No fiber to shore — it beams compute results back via low-Earth-orbit satellite

The raise (May 4, 2026):

· $140M Series B led by Peter Thiel (PayPal co-founder)
· Reported ~$1 billion valuation
· Joined by Fortescue Capital, Future Positive, Hanwha, Leblon Capital + 29 investors (~$218M raised total)
· Capital finishes a pilot manufacturing facility near Portland

The track record (not a pitch deck):

· Founded 2016 as a public benefit corporation
· Ocean-1, Ocean-2, and Wavehopper prototypes deployed in real sea trials in 2021 and 2024
· Ocean-3 pilot deploys in the northern Pacific this year; commercial systems in 2027
· 120 employees. CEO Garth Sheldon-Coulson is ex-Bridgewater

The trilogy — data centers where humans aren’t:

· SEA: Panthalassa — unlimited wave energy, no grid to fight
· SPACE: orbital data centers (our ORBIT post) — uninterrupted solar, radiative cooling
· LAND: Crusoe — compute co-located with stranded energy (flare gas, remote power)

· One thesis: when the grid can’t scale, take the data center to the fuel

Garth Sheldon-Coulson: “There are three sources of energy on the planet with tens of terawatts of new capacity potential: solar, nuclear, and the open ocean.”

— Frontier-first.

Sources: GeekWire · Fortune · DCD · BusinessWire · CB Insights · PitchBook · 05.21.26

#Panthalassa #AIInfrastructure #WavePower #DataCenters #VentureCapital

PANTHALASSA.

Floating AI data centers powered entirely by ocean waves. No grid. No anchors. Peter Thiel just led a $140M round at a reported $1 billion valuation.

The concept:

· Ocean-3 — an autonomous, self-propelled floating platform that generates clean wave energy to power a modular AI data center
· No anchors, no cables to land or seabed — a fully self-contained unit
· Power comes from waves: water forced through an internal turbine as the platform rises and falls, sited in the planet’s most energy-dense wave regions
· No fiber to shore — it beams compute results back via low-Earth-orbit satellite

The raise (May 4, 2026):

· $140M Series B led by Peter Thiel (PayPal co-founder)
· Reported ~$1 billion valuation
· Joined by Fortescue Capital, Future Positive, Hanwha, Leblon Capital + 29 investors (~$218M raised total)
· Capital finishes a pilot manufacturing facility near Portland

The track record (not a pitch deck):

· Founded 2016 as a public benefit corporation
· Ocean-1, Ocean-2, and Wavehopper prototypes deployed in real sea trials in 2021 and 2024
· Ocean-3 pilot deploys in the northern Pacific this year; commercial systems in 2027
· 120 employees. CEO Garth Sheldon-Coulson is ex-Bridgewater

The trilogy — data centers where humans aren’t:

· SEA: Panthalassa — unlimited wave energy, no grid to fight
· SPACE: orbital data centers (our ORBIT post) — uninterrupted solar, radiative cooling
· LAND: Crusoe — compute co-located with stranded energy (flare gas, remote power)

· One thesis: when the grid can’t scale, take the data center to the fuel

Garth Sheldon-Coulson: “There are three sources of energy on the planet with tens of terawatts of new capacity potential: solar, nuclear, and the open ocean.”

— Frontier-first.

Sources: GeekWire · Fortune · DCD · BusinessWire · CB Insights · PitchBook · 05.21.26

#Panthalassa #AIInfrastructure #WavePower #DataCenters #VentureCapital

.pro-gallery-wix-wrapper {display: block !important;} .pro-gallery-wix-wrapper .gallery-item-container {opacity: 1 !important; display: block !important;}

bottom of page