NNuggets
BookmarksCollections
  • About Us
  • Terms of use
  • Privacy policy
  • Disclaimer
  • Copyright & Takedown Policy
  • Community Guidelines
  • Cookie Policy
  • Contact

© 2026 Nuggets

NuggetsMarket PulseCollections

On this page

Early Breakthroughs

  • Early Breakthroughs
  • First Microchips
  • GPUs as AI Backbone
  • AI Model Training Acceleration

On this page

  • Early Breakthroughs
  • First Microchips
  • GPUs as AI Backbone
  • AI Model Training Acceleration
Technology/February 22, 2026/9 min read/ubs.com

Chips ahoy - The Semiconductor | 250 Years of US Innovation | Issue 5 | UBS

Source

"It was the government that created the large demand that facilitated mass production of the microchip." - Fred Kaplan, author of 1959: The Year Everything Changed (Discussing the role of the US Department of Defense in early microchip adoption)

"Before Silicon Valley became today's hub, or central processing unit, for technology, there were numerous foundational discoveries dotting the US landscape along the way." - UBS Authors (Reflecting on the geographically diverse origins of US tech innovation)

"The early semiconductor industry was highly reliant on federal government policy and financial support, especially for defense and aerospace, to achieve scale and ultimately realize resounding commercial success." - UBS Authors (Highlighting the crucial role of government funding in the sector's infancy)

"While there is much anticipation about the future for semiconductors, nearly 70 years after their initial discovery, there is also no room for complacency." - Ulrike Hoffmann-Burchardi (Concluding on the modern AI race and US manufacturing lags)


1. Executive Overview: Semiconductors as the Core of Modern Innovation

References

  1. Original source (ubs.com)

Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer

Related nuggets

Jun 2, 2026

AI Is Escaping the Screen | 01 Jun 2026 | Coatue

Coatue : AI is entering a new phase: moving beyond digital tools and into fully autonomous systems operating in the physical world. From advanced manufacturing and surgical robotics to robots in the home, the next wave of innovation will b…

Jun 2, 2026

Kalshi Monthly Volume - Politics ($M) | Chart of the Day | Coatue

Coatue: Kalshi's political volume has scaled dramatically, and the American Power Index KPOW is what that scale enables: a single number gauge of the current balance of political power and where markets expect it to move, which Kalshi bill…

Jun 2, 2026

The BlackBerry Problem |18 May 2026 | The Mistakes Series | Malcolm Gladwell's Revisionist History

"My mistake and naivity was to think that people are were with me so you're flying around the world you're trying to get people on side and you think they're on side but they're not mhm mhm and you get blindsight" Jim Balsillie 00:01:34 ht…

Jun 2, 2026

Partnership Perspectives: Network International | 2 Jun 2026 | Brookfield Perspectives

Actions

Reading

Published
February 22, 2026
Read time
9 min read
Progress0%

This report situates semiconductors at the center of 250 years of US innovation, arguing that microchips are foundational to modern economic growth, AI development, national security, and geopolitical competition.

Key thesis:

  • The semiconductor industry was initially catalyzed by US government demand (defense and aerospace).
  • It evolved into a commercial powerhouse through scaling, Moore’s Law, and globalization.
  • Today it stands at the center of the AI arms race, massive hyperscaler capital spending, and geopolitical rivalry (US–China).
  • The current AI infrastructure buildout may become the largest megaproject in human history.

2. Foundational Origins (1940s–1970s)

Early Breakthroughs

The mythology of Silicon Valley overlooks early national innovation hubs:

  • 1947 – Bell Labs: Invention of the transistor.
  • April 1954 – Texas Instruments (Gordon Teal): First commercially viable silicon transistor (superior to germanium for high-temperature military/industrial use).
  • 1957 – Jay Lathrop: Photolithography breakthrough (miniaturization via inverted microscope lens).
  • 1958 – Jack Kilby (TI): First integrated circuit.
  • 1959 – Robert Noyce (Fairchild Semiconductor): First practical silicon microchip suitable for mass production.
  • 1968 – Intel founded by Noyce and Gordon Moore.

First Microchips

  • Contained 4 transistors
  • Cost USD 31 in 1960 (~USD 330 in 2025 dollars)
  • Used in:
    • US Air Force Minuteman II missile systems
    • NASA Apollo missions

As shown in Figure 1 (page 5):

  • US integrated circuit sales were initially military-dominated.
  • Commercial sales eventually overtook military demand.

3. Moore’s Law and Exponential Scaling

Moore’s Law (Gordon Moore): Transistors double every ~2 years.

From Figure 2 (page 5):

  • Process nodes shrank from thousands of nanometers (1970s) to a few nanometers today.
  • Producer price index for integrated circuits fell sharply over decades.
  • Early 1970s: US Department of Defense mass production reduced chip costs to USD 1.25 per unit.

Fred Kaplan quote:

“It was the government that created the large demand that facilitated mass production of the microchip.”


4. Productivity Explosion (1970s–1980s)

From Figure 3 (page 6):

  • Semiconductor industry productivity growth (1972–1986):
    • +13% annualized
  • Overall US productivity:
    • ~2%

Output per hour rose more than twice as fast as employment.

This period marked:

  • Falling chip prices
  • Expanding adoption (calculators, hearing aids, PCs)
  • Early digital revolution

5. Japan Shock and Trade Intervention (1980s)

From Laura D’Andrea Tyson (1992):

DRAM market share shift (1978–1986):

  • US: 70% → 20%
  • Japan: <30% → ~75%

1986 US–Japan Semiconductor Agreement

  • Limited Japanese exports
  • Effectively fixed chip prices
  • Preserved US leadership for two decades
  • Resulted in higher prices for buyers

Afterward, US shifted focus toward:

  • Chip design
  • R&D
  • Outsourced manufacturing

6. Taiwan’s Strategic Rise

  • 1980s: Hsinchu Science Park established
  • Industrial policy:
    • Heavy R&D investment
    • Recruitment of US-trained engineers
  • Result: Taiwan becomes global manufacturing powerhouse

7. Long-Term Equity Performance

From Figure 4 (page 6):

Semiconductor stocks:

  • Outperformed in 1970s
  • Underperformed late 1980s (DRAM collapse, capital intensity)
  • Strong resurgence in 1990s onward
  • Long-term structural outperformers vs. tech and S&P 500

8. Infographic: Scale of Technological Transformation (Page 3)

Key data comparisons:

Metric19612025
Transistors416.6 billion
RAM4 kB12.6 million kB
CPU speed100–300 kHz4,300,000 kHz
Price per chipUSD 335 (1961)USD 500 (2025)

Apollo-era computer vs. modern smartphone illustrates exponential gains.

Additional data:

  • US top exports 2024:
    • Semiconductors: USD 57bn
    • Natural gas: USD 62.6bn
    • Cars: USD 61bn
    • Aircraft: USD 123.3bn
    • Refined oil: USD 123.4bn

Quantum comparison:

  • Google’s Willow quantum chip: <5 minutes
  • Equivalent classical compute: 10 trillion years

9. The AI Era: GPUs and Specialized Chips

GPUs as AI Backbone

Modern GPUs:

  • Thousands of stream processing cores
  • High-bandwidth memory (HBM)
  • Advanced interconnects

Operational economics (Figure 5, page 8):

  • Estimated operational cost: ~$0.40 per GPU per hour
  • Rental prices significantly higher

AI Model Training Acceleration

From Figure 6:

  • AI training compute doubling every ~6 months
  • Notable models: AlexNet → GPT-3 → GPT-4 → Gemini Ultra → Grok 4
  • 4.4x acceleration slope indicated

10. Advanced Packaging Revolution

Three key technologies:

1. CoWoS (Chip-on-Wafer-on-Substrate)

  • 2.5D packaging
  • Silicon interposer
  • Enables dense logic-HBM connections
  • Gold standard for AI GPUs

2. CoPoS (Chip-on-Panel-on-Substrate)

  • Panel-based (not wafer-based)
  • Larger rectangular substrates
  • Supports massive “superchips”
  • Reduces material waste

3. CPO (Co-Packaged Optics)

  • Fiber optics integrated at chip level
  • Speeds of 1.6 terabits+
  • Reduces latency and heat
  • Critical for next-gen data centers

11. AI Compute Capacity and Market Concentration

From Figure 7 (page 9):

  • AI compute capacity increasing ~3.4x per year since 2022
  • Doubling roughly every 7 months
  • NVIDIA = >60% of total compute
  • Google & Amazon = majority of remainder

Power distribution insight:

  • GPUs = ~40% of peak power consumption
  • Majority of energy used by cooling, inefficiencies, interconnects

12. Geopolitical Positioning

From Figure 8:

Global GPU cluster performance share (2025):

  • US: ~75%
  • China: ~15%
  • Others (EU, Japan, Norway, etc.): smaller shares

China:

  • Domestic AI chips (Huawei Ascend series)
  • State-backed capex
  • Estimated localization rate:
    • ~40% domestically made GPUs by 2027

US:

  • CHIPS and Science Act (2022)
  • Export controls on advanced AI chips
  • Goal: regain manufacturing edge

Semiconductors framed as:

  • National security asset
  • Economic sovereignty pillar
  • Geopolitical leverage tool

13. Hyperscaler Capex Explosion

Q4 2025 earnings season:

Alphabet, Meta, Amazon:

  • Capex ~30% above initial consensus

Top five hyperscalers (incl. Microsoft, Oracle):

  • CY 2026 capex: USD 630bn
  • CY 2027 capex: USD 729bn
  • Upward revisions:
    • +18% (2026)
    • +20% (2027)

Free Cash Flow impact:

  • 2026 FCF: USD 91bn (–49% vs YE 2025 estimates)
  • 2027 FCF: USD 149bn (–38%)

Meta may become net debt positive if capex exceeds internal cash generation.


14. 2030 Capex Projection and Productivity Threshold

UBS estimate:

  • AI capex could reach USD 1.3 trillion by 2030
  • Implies 25% CAGR (2025–2030)

To justify:

  • ~USD 6 trillion productivity gains by 2030
  • Equivalent to ~10% of today’s global labor market

UBS view: achievable based on historical productivity trends.


15. Vendor Financing Risks (Dotcom Echo)

Dotcom peak:

  • Vendor financing >120% of pretax earnings

Today:

  • NVIDIA vendor-linked collaborations ≈ 10% of 2026 pretax earnings

Risk present but materially smaller than 2000 bubble conditions.


16. Structural Conclusion

The semiconductor industry has evolved from:

  • Government-supported defense technology
    → Mass commercial adoption
    → Globalized supply chains
    → AI megaproject infrastructure backbone

Key framing:

  • The AI buildout may become the largest megaproject in human history
  • Transition toward:
    • Artificial General Intelligence (AGI)
    • Token-based digital economy
  • AI tokens as fundamental economic unit of compute and intelligence
  • Semiconductors positioned as core driver of economic value capture

UBS frames this within its: AI Transformational Innovation Opportunity (TRIO) thesis.


17. Strategic Implications

  1. US leads in AI compute dominance but lags in manufacturing.
  2. China accelerating localization (~40% domestic GPUs by 2027).
  3. Hyperscaler capex peaking but structurally expanding.
  4. AI productivity gains must justify trillion-dollar infrastructure.
  5. Chips are now inseparable from:
    • National security
    • Energy infrastructure
    • Cloud economics
    • Labor productivity
    • Digital asset ecosystems

Fascinating Facts & Infographic Statistics

  • Transistor Growth: Transistor counts per chip evolved from 4 in 1961, to 2,300 in 1971, to 1.35 million in 1994, to an astounding 16.6 billion in 2025.
  • CPU Speed: Speeds have surged from 100-300 kilohertz in 1961 to 4.3 million kilohertz in 2025.
  • Top US Exports in 2024: Semiconductors ($57bn), Cars ($61bn), Natural gas ($62.6bn), Aircraft ($123.3bn), and Refined oil ($123.4bn).
  • Quantum Leap: A standard computational benchmark that would take today's fastest supercomputers 10 trillion years to complete takes Google's Willow (Quantum chip) less than 5 minutes.
  • Miniaturization Perspective: The size of a microchip node in 1971 was 10,000 nanometers; if transistor sizes had stayed constant from 1971, a 2025 microchip would measure 3,824,900,000 nanometers.
  • Memory Capacity: The computer used for the Apollo moon missions possessed 4 KB of RAM and 73.7 kB of ROM, whereas modern smartphones boast 12.6 million kB of RAM and 268.4 million kB of ROM.
  • Chemical Complexity: The number of periodic table elements used to manufacture a semiconductor increased from 17 in the 1980s, to 21 in the 1990s, to 62 in the 2000s, to 66 in the 2010s.

Stories & Anecdotes

  • The Apollo Mission's Modest Computing: To illustrate the staggering pace of innovation, the report contrasts the computer that guided the Apollo missions to the moon with a modern smartphone. The Apollo computer operated on a mere 4 KB of RAM and 73.7 kB of ROM. Today, a standard smartphone carries roughly 12.6 million kB of RAM, holding exponentially more power in a consumer's pocket than what NASA used to reach space.
  • The Minuteman II & Mass Production: Early microchips were incredibly expensive. The narrative details how the US Department of Defense stepped in as the ultimate "early adopter." By procuring these chips for the Minuteman II missile guidance systems, the government effectively subsidized the creation of mass production lines, ultimately driving consumer prices down from $31 to $1.25 and launching the digital revolution.
  • The Scale of Miniaturization: The report provides a striking hypothetical to explain miniaturization: The size of a microchip node in 1971 was 10,000 nanometers. If transistor sizes had stayed constant from 1971, a 2025 microchip would measure a staggering 3,824,900,000 nanometers to achieve current capabilities.

References & Recommendations

Books:

  • 1959: The Year Everything Changed, Fred Kaplan - Quoted to emphasize how early US government demand created the mass production necessary for the microchip's success.

People Referenced:

  • Gordon Teal - Texas Instruments team leader who commercialized the first silicon transistor in 1954, paving the way for military and industrial uses.
  • Jay Lathrop - Advanced the field in 1957 by developing photolithography using an inverted microscope lens, a critical step for miniaturization.
  • Jack Kilby - Engineer at Texas Instruments who successfully integrated all electronic circuit components onto a single piece of semiconductor material in 1958, inventing the first microchip.
  • Robert Noyce & Gordon Moore - Engineers at Fairchild Semiconductor who filed the patent for the first practical silicon microchip in 1959 and later founded Intel Corporation in 1968.

Tools/Platforms/Products:

  • Intel 4004 - Released in 1971, noted as one of the first commercially available microprocessors that integrated an entire CPU onto a single silicon chip.
  • Google Willow - A modern quantum chip referenced for completing a standard computational benchmark in under 5 minutes, a task that would take a traditional supercomputer 10 trillion years.

Speakers & Credentials

The report is authored and compiled by the Chief Investment Office at UBS Wealth Management, featuring contributions from senior leadership:

  • Ulrike Hoffmann-Burchardi - Chief Investment Officer Americas and Global Head of Equities, UBS. (Provides strategic oversight and concluding remarks on the future of US semiconductor capabilities).
  • Kurt Reiman - Head of Fixed Income CIO Americas & Editor-in-Chief.
  • Delwin Limas & Kayden Lee - Equity Strategists, CIO APAC.

"Brookfield's the largest infrastructure owner in the world... We drew a pipeline and we showed all the different components of the payments ecosystem on a pipeline and said it's like a pipe that moves any commodity except what it's moving…