Hybrid and Post-Quantum Computing
Lessons from Dot-Com, Blockchain, the Metaverse and What Practitioners Should Do Now
Yes, there are bubble dynamics in today’s AI boom, overheated valuations, vast capex, and plenty of “pilot-itis.” But like the internet after 2000, serious development will continue and compound. The next 12–24 months are likely to separate hype from durable value.
What we mean by “a bubble” (and why AI feels like one)
A bubble isn’t just high prices; it’s expectations running ahead of fundamentals. We’ve seen the pattern before:
- Dot-com (1995–2001): massive capital, weak unit economics; crash, then two decades of real value creation (search, cloud, mobile).
- Blockchain (2017–): breakthrough primitives, but enterprise value concentrated in narrow rails while many promises stalled.
- Metaverse (2021–): genuine tech progress, but consumer/enterprise pull proved shallower and slower than forecasts.
AI shows the same installation phase traits: outsized infrastructure spends, fast-moving narratives, and unclear near-term Return of Investment (ROI) for many adopters. Analysts now track data-centre capex as a core macro indicator; some estimates foresee hundreds of billions across hyperscalers in 2024–2025 and multi-trillion totals over the decade, classic signs of an early “build-out” period before returns settle.
At the same time, public-market concentration risk is real. Nvidia’s prints and guideposts have become a market-wide AI barometer; even when the numbers are strong, valuation sensitivity is extreme, another bubble hallmark.
The AI moment by the numbers (as of September 2025)
- Adoption: Enterprise use of AI (including generative AI) climbed sharply through 2024; McKinsey reported a step-change to ~70%+ adoption, with revenue impact rising in units that deploy seriously.
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2024
- Productivity potential: Goldman Sachs’ latest read still pegs long-run labour-productivity uplift in developed markets at ~15% from genAI, with transitional labour frictions. Earlier work suggested ~7% uplift to global GDP over a decade.
https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-global-workforce
- Hype vs. fundamentals: Gartner’s Hype Cycles place many genAI components on the steep slope of inflated expectations, helpful context when boards extrapolate too fast.
https://www.gartner.com/en/articles/hype-cycle-for-genai
- Regulatory gravity: The EU AI Act is now law, with risk-tiered obligations and enforcement timelines that will force governance, documentation, and transparency: the AI analogue of GDPR’s discipline on data
Where returns are showing and where they’re not (yet)
- Emerging bright spots: Decision-support in knowledge work, customer operations, coding assistance, and summarisation/search are yielding measurable gains in the units that implement well. Multiple 2024–2025 surveys show positive ROI for a meaningful subset of adopters, especially where foundations (data quality, security, change management) are in place.
https://www.forrester.com/report/areas-of-positive-roi-from-generative-ai-are-now-on-par-with-predictive-ai/RES181712
- The J-curve reality: As with past general-purpose technologies, measured productivity often lags during the build-out of intangible complements (data pipelines, process redesign, skills). Expect a dip before the harvest, the Productivity J-curve.
https://ide.mit.edu/sites/default/files/publications/2019-04JCurvebrief.final2_.pdf
- The ROI gap: A large share of firms still struggle to scale beyond pilots; breadth beats depth in slideware, but depth beats breadth in P&L. BCG’s 2024 work found ~three-quarters of companies weren’t yet achieving or scaling value; leaders narrowed scope to a handful of high-impact use cases.
What could actually “pop” the bubble?
-
Capex-to-cash conversion risk: If infra spend outpaces monetisation for too long, markets will re-rate the AI complex (chips, cloud, model providers). Watch hyperscaler capex guides, GPU supply chains, and power constraints.
-
Unit economics of inference: Inference costs must fall faster than usage grows for many consumer and enterprise models to sustain margins. If they don’t, volume caps and pricing backlash can bite. (Industry trackers and provider blogs increasingly emphasise cost-efficiency.)
https://learn.microsoft.com/en-us/azure/ai-foundry/openai/whats-new - Governance and compliance drag: The EU AI Act and adjacent frameworks (copyright transparency, high-risk controls) will raise the bar for documentation, testing, and incident handling rightly so. Firms without governance patterns will slow or pause roll-outs. https://artificialintelligenceact.eu/high-level-summary/
- Mismatch between “agent” promise and production reality: Agentic systems are genuinely promising for long-running, multi-step tasks, but only when bounded, well-tooled, and supervised. Microsoft, Google and others are productising agents; results vary, and governance is essential. https://www.theverge.com/google-io/670386/google-astra-universal-ai-assistant-prototype-io-2025
Three plausible 12–24-month scenarios
- Pop & Reset (shallow recession in AI equities): Infra leaders survive; frothier apps consolidate. The installation phase rolls into a more disciplined deployment phase with efficiency overgrowth.
https://www.ft.com/content/a76f238d-5543-4c01-9419-52aaf352dc23 - Soft Deflation: Valuations cool without a shock; cost curves and tooling maturity improve; governance frameworks standardise; ROI widens from “AI-first” to “AI-capable” firms.
- Diffusion & Durable Growth: Rapid agentic automation plus cheaper inference create step-function time-savings in back-office and sales ops; productivity effects begin to show up in national stats.
- Leading indicators to watch: hyperscaler capex vs. AI service revenue; GPU/TPU utilisation; power availability; inference pricing trendlines; EU AI Act compliance milestones; enterprise deployment of agent capabilities in Microsoft 365, Dynamics, and Google Workspace.
A pragmatic playbook (what to actually do in Q4/25)
A) Pick fewer, bigger use cases. Treat the portfolio like a barbell: 2–3 near-term “timeback” bets (email triage, meeting actions mining, sales support from CRM data) and 1–2 mid-term process automations (claims, employee onboarding). Leaders focus rather than spray-and-pray.
B) Put your data house in order: You don’t need a perfect lakehouse, but you do need governed sources of truth, lineage, access policies, and retrieval patterns (RAG) that avoid brittle prompt spaghetti.
C) Engineer for cost & safety.
- Track cost per task for both training (if relevant) and inference.
- Implement guardrails (prompt validation, structured outputs, human-in-the-loop) and log everything for auditability.
https://azure.microsoft.com/en-us/blog/announcing-a-new-openai-feature-for-developers-on-azure/
D) Standardise on the platforms your people already use. For knowledge work, that increasingly means Microsoft 365 + Copilot, extended with Copilot Studio agents for long-running, multi-step processes (e.g., triage → draft → route → follow-up). Start with narrow autonomy and clear stop conditions.
E) Measure like finance, not like labs. Define hard baselines (cycle time, tickets closed per Full-Time Equivalent (FTE), renewal rate, pipeline velocity) and compare pre/post. If it doesn’t move a Key Performance Indicator (KPI), it’s a demo — not a deploymen
Where Asygma fits
We’re not here to sell fairy dust; we bring AI inside your Microsoft estate, private, compliant, and secure and we operationalise it:
- Timeback design sprints: Identify 2–3 high-leverage use cases and build the KPI tree behind them.
- Copilot inside M365, with governance: Roll out safely with Zero-Trust controls and monitoring.
- Agentic automations: Build Copilot Studio agents that execute repeatable workflows (with logs, approvals, and cost caps).
https://www.microsoft.com/en-us/microsoft-365/blog/2024/05/21/new-agent-capabilities-in-microsoft-copilot-unlock-business-value/
- Change & adoption: Training, comms, and “show-me” playbooks so people actually use what we deploy.
Don’t wait for the bubble to burst to find out what works
Asygma helps organisations move from pilots to real outcomes, deploying Microsoft 365 Copilot and AI agents with governance, adoption and ROI in mind.
References
BCG, 2024. AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value. [online] Available at: https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scale-value [Accessed 31 August 2025].
Carvao, P., 2025. Is the AI Bubble Bursting? Lessons From the Dot-Com Era. Forbes, [online] 21 August. Available at: https://www.forbes.com/sites/paulocarvao/2025/08/21/is-the-ai-bubble-bursting-lessons-from-the-dot-com-era [Accessed 1 September 2025].
European Commission, 2025. EU Artificial Intelligence Act. [online] Available at: https://artificialintelligenceact.eu/the-act [Accessed 1 September 2025].
Gartner, 2024. Hype Cycle for Artificial Intelligence 2024/2025. [online] Available at: https://www.gartner.com/en/research/methodologies/gartner-hype-cycle [Accessed 4 September 2025].
Goldman Sachs, 2023. The Potential Long-Run Impact of Generative AI on Productivity. [online]. Available at: https://www.goldmansachs.com/insights/pages/generative-ai-could-raise-global-gdp-by-7-percent.html [Accessed 4 September 2025].
Harvard Business Review, 2025. The AI Revolution Won’t Happen Overnight. [online] 24 June. Available at: https://hbr.org/2025/06/the-ai-revolution-wont-happen-overnight [Accessed 31 August 2025].
Investopedia, 2025. Are AI Stocks the Next Pets.com? Key Bubble Indicators Investors Should Watch. [online] August. Available at: https://www.investopedia.com/ai-stocks-and-signs-of-a-bubble-11765638 [Accessed 4 September 2025].
McKinsey, 2024. The state of AI in early 2024: Gen AI adoption spikes and starts to generate value. [online] Available at: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2024 [Accessed 4 September 2025].
Microsoft, 2025. Introducing agent flows: Transforming automation with AI-first workflows. [online] 02 April. Available at: https://www.microsoft.com/en-us/microsoft-copilot/blog/copilot-studio/introducing-agent-flows-transforming-automation-with-ai-first-workflows [Accessed August 30 2025].
MoneyWeek, 2025. Could the AI megacap bubble burst? [online] August. Available at: https://moneyweek.com/investments/tech-stocks/could-ai-megacap-bubble-burst [Accessed 4 September 2025].
New York Post, 2025. OpenAI CEO Sam Altman warns of AI bubble, says investors are “overexcited”. [online] 18 August. Available at: https://nypost.com/2025/08/18/business/openai-ceo-sam-altman-warns-of-ai-bubble-says-investors-are-overexcited-report [Accessed 4 September 2025].
Sharma, A., 2025. OpenAI CEO Sam Altman warns startups: “Someone’s gonna get burned there”. Times of India, [online] August. Available at: https://timesofindia.indiatimes.com/technology/tech-news/someones-gonna-get-burned-there-with-these-crazy-openai-ceo-sam-altman-warns-startups/articleshow/123356050.cms [Accessed 4 September 2025].
Tom’s Hardware, 2025. AI bubble is worse than the dot-com crash that erased trillions, economist warns. [online] July. Available at: https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-bubble-is-worse-than-the-dot-com-crash-that-erased-trillions-economist-warns-overvaluations-could-lead-to-catastrophic-consequences [Accessed 4 September 2025].
Wall Street Journal, 2025. What the Dot-Com Bust Can Tell Us About Today’s AI Boom. [online] Available at: https://www.wsj.com/tech/ai/what-the-dot-com-bust-can-tell-us-about-todays-ai-boom-c78482e7 [Accessed 29 August 2025].
White & Case, 2024. Long awaited EU AI Act becomes law after publication in the EU’s Official Journal. [online] Available at: https://www.whitecase.com/insight-alert/long-awaited-eu-ai-act-becomes-law-after-publication-eus-official-journal https://www.whitecase.com/insight-alert/eu-ai-act-what-companies-need-know[Accessed 4 September 2025].
Wikipedia, 2025. AI bubble. [online] Available at: https://en.wikipedia.org/wiki/AI_bubble [Accessed 4 September 2025].
Wikipedia, [n.d.]. Dot-com bubble. [online] Available at: https://en.wikipedia.org/wiki/Dot-com_bubble [Accessed 4 September 2025].
Windows Central, 2025. The AI bubble may be about to pop – here’s what MIT’s 95% failure stat means. [online] August. Available at: https://www.windowscentral.com/artificial-intelligence/the-ai-bubble-may-be-about-to-pop-heres-what-mits-95-percent-failure-stat-means [Accessed 4 September 2025].
Xu, Y. et al., 2022. Fusing Blockchain and AI with Metaverse: A Survey. arXiv, [online] Available at: https://arxiv.org/pdf/2201.03201 [Accessed 4 September 2025].
Xu, Y. et al., 2022. Cryptocurrency bubbles, the wealth effect, and non-fungible token prices: Evidence from metaverse LAND. arXiv, [online] Available at: https://arxiv.org/pdf/2209.04385 [Accessed 4 September 2025].
Zhang, C. et al., 2022. AI and 6G into the Metaverse: Fundamentals, Challenges and Future Research Trends. arXiv, [online] Available at: https://arxiv.org/pdf/2208.10921 [Accessed 4 September 2025].
Zhao, Y. et al., 2024. Dot-com and AI bubbles: Can data from the past be helpful? ScienceDirect, abstract [online] Available at: https://www.sciencedirect.com/science/article/abs/pii/S1544612324008298 [Accessed 4 September 2025].
Zhao, Y. et al., 2024. Is the Metaverse Dead? Insights from Financial Bubble Theory. MDPI, [online] Available at: https://www.mdpi.com/2674-1032/3/2/17 [Accessed 4 September 2025].
Zhu, X. et al., 2025. Anatomy of a Digital Bubble: Lessons Learned from the NFT and Metaverse Frenzy. arXiv, [online] 16 January. Available at: https://arxiv.org/pdf/2501.09601 [Accessed 4 September 2025].
Glossary (fast, practitioner-friendly)
AI bubble: A period when expectations and valuations for AI assets run ahead of monetisable fundamentals.
Agent / Agentic AI: Systems that can plan and execute multi-step tasks on your behalf (with tools, memory, and supervision). In Microsoft 365, agents are built via Copilot Studio.
Capex vs. Opex (AI): Capital spend (chips, data centres, power) vs. operating costs (inference per request, recurring, orchestration). Markets watch how capex converts to cash.
EU AI Act (risk-based regulation): Europe’s binding framework that bans “unacceptable-risk” uses, regulates “high-risk” systems, and imposes transparency on “limited-risk” use cases. Staged enforcement through 2026.
Inference: Running a model to get outputs; the recurring cost that dominates production unit economics.
J-curve (Productivity): The empirically observed lag between investing in a new general-purpose technology and when measured productivity rises.
https://www.aeaweb.org/articles?id=10.1257%2Fmac.20180386
Pilot-itis: refers to the phenomenon originated in the healthcare industry where there is a tendency to launch numerous small-scale pilot projects without committing to or scaling them to full-scale implementations.
RAG (Retrieval-Augmented Generation): Pattern that grounds model outputs in your approved data sources to reduce hallucinations and improve accuracy.
