Even Before AI, Data Center "Net Zero" Was an Impossible Goal, for Most

It is in some ways refreshing when realistic assessments are made about potential progress towards climate goals, as we are unlikely to meet the current goals in any case, as Eric Schmidt, former Google CEO, notes. “We're not going to hit the climate goals anyway,” he bluntly said. 

We might as well be realistic about how much can be done, in terms of carbon and other emissions. Nearly every stated goal we see is unlikely to be met. We might not prefer that outcome, but that seems to be the reality, so we might as well work on what can be accomplished, rather than pursuing unworkable plans that already are destined to fail.

And that applies to data center sustainability as to all other sustainability goals. In other words, present net zero goals, to say nothing of negative emissions profiles, are impossible in the artificial intelligence era. 

Of course, any person, company or nation can choose not to use AI, but that will not prevent climate goal failure. That would happen even in the absence of any additional AI processing activities.  

Climate Target	Organization/Agreement	Target Deadline	Goal	Current Progress and Likelihood of Success
Net-Zero Global Carbon Emissions	Paris Agreement (2015)	2050	Limit global warming to below 2°C, preferably 1.5°C above pre-industrial levels	Unlikely under current policies. Global emissions are still rising, and only a few countries have detailed roadmaps to achieve net-zero by 2050. A massive scale-up in renewable energy, electrification, and industrial reform is required.
Reduce Emissions by 45% from 2010 levels	Paris Agreement (IPCC SR15)	2030	Achieve 1.5°C temperature increase limit	Unlikely. While some regions (EU) have committed to reductions, global emissions need to fall by 7.6% annually to meet this target, but they are still rising.
Phase out coal power plants	Various pledges, COP26	2030s (developed); 2040s (developing)	End reliance on coal as a primary energy source	Mixed. Coal usage has been reduced in the EU and US, but major emitters like China and India continue to expand coal power. Global coal phase-out is slow.
Methane emissions reduction by 30% from 2020 levels	Global Methane Pledge (COP26)	2030	Reduce methane emissions to limit short-term warming	Uncertain. Methane emissions from agriculture and oil/gas remain high, but new policies to detect and reduce leaks in energy and industry are promising.
100% sales of zero-emission vehicles (ZEVs)	COP26 ZEV Declaration	2040 globally; 2035 for major markets	End sales of internal combustion vehicles	Possible in some regions. Several countries (UK, EU) have set ambitious bans on fossil-fuel vehicles, but progress in developing markets is slower.
Halting deforestation and land degradation	Glasgow Leaders’ Declaration (COP26)	2030	Stop deforestation to protect carbon sinks	Unlikely. Deforestation continues in regions like the Amazon and Congo. Efforts to reduce illegal logging are hindered by political and economic factors.
Finance for climate mitigation and adaptation	Paris Agreement and UNFCCC	$100 billion per year by 2020, extended through 2025	Provide climate financing for developing countries	Behind schedule. Developed nations have not fully met this target. The annual climate finance commitment remains short of the promised $100 billion.
Energy from renewable sources at 60-80%	Various (IEA, IPCC reports)	2050	Global energy generation from renewables	Possible but challenging. The renewable energy capacity is growing rapidly (solar, wind), but fossil fuels still dominate energy consumption globally.
Achieving Circular Economy principles	EU Circular Economy Action Plan, UN SDGs	2030	Reduce waste, increase recycling and efficiency	Possible in some regions. The EU leads in circular economy initiatives, but widespread adoption globally is still limited by lack of infrastructure and policy.
Ocean Protection: 30x30 Initiative	High Ambition Coalition (Biodiversity)	2030	Protect 30% of the world's oceans	Uncertain. Ocean protection commitments are increasing, but enforcement of protected areas and addressing overfishing remain major challenges.
Climate Resilient Infrastructure	Paris Agreement, IPCC	Ongoing	Build resilient infrastructure to withstand climate impacts	Lagging. Many countries lack adequate infrastructure planning for climate resilience. Investments in adaptation are growing but still insufficient.

Yes, Virginia, You Can Yell "Fire" in a Crowded Theater

As it turns out, one actually can lawfully “yell ‘fire’ in a crowded theater,” the traditional example of a limitation of free speech protections under the First Amendment to the U.S. Constitution.

When U.S. politicians support arguments for greater censorship, they often argue that such First Amendment rights are restricted. The common refrain is that one cannot “yell ‘fire’ in a crowded theater” or that “misinformation” or “hate speech” are similarly not protected.

But many would argue such interpretations are indeed calls for restrictions (violations) of the First Amendment to the U.S. Constitution. Since the time of the Supreme Court’s Schenck v. U.S. decision, where the phrase “shouting ‘fire’ in a crowded theater first emerged, courts have ruled that incendiary, distasteful, rude speech actually is protected.

Perhaps the clearest clarification of Schenck is Brandenburg. Unless immediate illegal action is intended and likely, even “yelling ‘fire’ in a crowded theater” is protected speech. 

The other obvious problem is that “misinformation” or “hate speech,” even if odious, inflammatory or believed to be false, still is protected speech. Ideas “one hates” remain the test of free speech protections. 

Case	Year	Ruling Summary	Relation to 'Yelling Fire' Metaphor
Schenck v. United States	1919	Upheld conviction for distributing anti-draft pamphlets during WWI. Established that speech presenting a "clear and present danger" could be restricted.	Justice Holmes introduced the famous metaphor of falsely shouting "fire" in a crowded theater.
Abrams v. United States	1919	Upheld convictions under the Espionage Act for distributing anti-war leaflets.	Built upon the "clear and present danger" standard, but Holmes dissented, moving towards more speech protection.
Gitlow v. New York	1925	Upheld conviction of a socialist for advocating the violent overthrow of the government.	Suggested that even speech not directly causing harm could be limited if it had the potential to incite violence.
Dennis v. United States	1951	Upheld convictions of communist leaders advocating for the violent overthrow of the U.S. government.	Argued that advocating dangerous ideas, even without immediate action, could be restricted.
Yates v. United States	1957	Ruled that advocating abstract doctrine (such as communism) is protected speech, unless it incites illegal action.	Limited previous rulings, clarifying that abstract ideas are protected unless linked to action.
Brandenburg v. Ohio	1969	Overturned conviction of a KKK leader for inflammatory speech. Held that speech is protected unless it incites imminent lawless action.	Effectively replaced the "fire in a crowded theater" metaphor. Protected even dangerous speech unless immediate illegal action is intended and likely.
Texas v. Johnson	1989	Ruled that flag burning constitutes protected free speech under the First Amendment.	Protected highly offensive speech, further distancing from "fire in a theater" metaphor.
Snyder v. Phelps	2011	Protected Westboro Baptist Church's right to protest at military funerals, ruling that offensive speech on public issues is protected.	Even offensive and distressing speech was deemed protected. Shows broad protection for public speech.

The phrase "freedom for the idea one hates" is closely associated with Justice Oliver Wendell Holmes and Justice Louis Brandeis, two influential figures in the development of free speech jurisprudence in the early 20th century. 

In his famous dissent in United States v. Schwimmer (1929), Holmes said "If there is any principle of the Constitution that more imperatively calls for attachment than any other, it is the principle of free thought—not free thought for those who agree with us but freedom for the thought that we hate."

The point is that free speech requires freedom for “highly offensive” or “controversial” speech. Without tolerance for such speech, the government could censor any speech that threatens its interests.

The phrase "Yes, Virginia" The phrase "Yes, Virginia" comes from a famous editorial published in The New York Sun on September 21, 1897, in response to a letter from an 8-year-old girl named Virginia O'Hanlon. Virginia had written to the newspaper asking whether Santa Claus really existed, because some of her friends had told her he did not.

Her letter read:

"Dear Editor: I am 8 years old.
Some of my little friends say there is no Santa Claus.
Papa says, 'If you see it in The Sun, it’s so.'
Please tell me the truth, is there a Santa Claus?"

The editorial was written by Francis Pharcellus Church, a veteran journalist. His response, titled "Yes, Virginia, there is a Santa Claus," became one of the most famous newspaper editorials in American history.

Why Marginal Cost of Content Creation is Generative AI's Superpower

Virtually every observer might agree that artificial intelligence will automate laborious tasks and therefore increase process efficiency. AI should also accelerate decision making, as it enables rapid information processing. 

podcast of this content

AI should enable more personalization than already is possible for user interactions and experiences and as a byproduct could change the nature of work, entertainment and learning. 

Generative AI, though, might bring cost impact in different ways than did other computing innovations. Virtually all computing eras since the advent of the personal computer have led to lower marginal costs of doing things. 

PCs meant computing power itself was widely available to people. The internet attacked the cost of sharing information and communicating while cloud computing arguably reduced software distribution costs while boosting the ability to apply accumulated data and insights more widely in real time. 

The mobile era extended computing capabilities “everywhere” and untethered from desks, tables or laps. 

Era	Computing Paradigm	Marginal Cost Implications
PC	Personal Computing	- High upfront costs for hardware and software - Relatively high marginal costs for upgrades and maintenance - Limited scalability
Internet	Networked Computing	- Reduced costs for information sharing and communication - Increased accessibility, but still significant infrastructure costs - Marginal costs tied to bandwidth and server capacity
Cloud Computing	On-Demand Computing	- Significantly lower upfront costs - Pay-as-you-go model reduces marginal costs - Improved scalability and flexibility - Potential for cost optimization through resource management1
Mobile	Ubiquitous Computing	- Lower device costs compared to PCs - App-based ecosystem with low distribution costs - Increased connectivity, but data costs can be significant
Future AI	Intelligent Computing	- Potential for near-zero marginal costs in some applications - High initial investment in AI development and infrastructure - Continuous learning and improvement may reduce long-term costs2

So it is reasonable to ask what the AI impact will be, especially generative AI, which seems to be driving mass market and most business AI use cases. 

Angela Strange, Andreessen Horowitz general partner and James da Costa Andreessen Horowitz partner, specialized in enterprise and business-to-business software, including financial technology. 

They believe the AI era leads to lower marginal cost of client and customer interactions, using AI agents to reduce the cost of labor involved in many customer support operations, including those involving information retrieval (files, ledger entries, past transitions, billing and account status). 

source: Andreessen Horowitz 

As applied in many areas outside of financial technology, the value of generative AI is squarely on its impact on content creation. 

Whether we look at text, image, video or audio, GenAI seems destined to have the highest impact on any process or industry built on content creation and its distribution or consumption. GenAI will be useful in any number of customer support contexts, but might be impactful in financial terms for the production of software and code; entertainment content; education and training; business communications; many types of research; marketing and sales.