Meta Reviews 2030 Net Zero Goal as Emissions Rise 45%

Meta is reassessing the timeline and interim targets for its 2030 net-zero goal as it expands AI infrastructure. Its 2026 sustainability reporting puts 2025 greenhouse-gas emissions at 11.8 million metric tonnes, up 45% year over year, while electricity use rose 35%. Meta attributes most of the emissions increase to its supply chain, including equipment and data-centre construction. It says it remains committed to net zero by 2030, is using renewable energy and improving data-centre efficiency, and has contracted for more than 30 gigawatts of clean power; about 12 gigawatts is currently generating electricity. The company says it may turn to natural-gas plants when utilities cannot supply power quickly enough.

AI-related security pressure is also growing. Microsoft CEO Satya Nadella warned that AI can uncover software vulnerabilities faster than organizations can safely test and deploy fixes. The Center for Internet Security said vulnerability disclosures rose from about 4,274 in October 2025 to 17,060 in October 2026, although roughly 90% are not practically exploitable. State and local governments face a growing patch burden, and recent Microsoft outages affecting Outlook, SharePoint and Teams underline the value of backup communications plans. CIS also urges trusted security practices and observed social engineering and typo-squatted domains during the disruptions.

Investment and research coverage points to both opportunities and costs in the AI build-out. Saturna Capital analyst Levi Stewart Zurbrugg highlighted Taiwan’s semiconductor industry, Dutch lithography, Korean memory makers and Japanese optical-fibre suppliers. He warned that hyperscalers’ depreciation costs could weigh on S&P 500 earnings growth in 2027 and 2028, and said investors should watch corporate debt and credit conditions as companies use more outside financing. Saturna’s Islamic screening limits exposure to companies whose debt exceeds 30% of market capitalisation; Zurbrugg also discussed diversification through the Saturna Global Equity ETF.

In AI research, the TailWorld benchmark tests whether models can infer physical rules from video and apply them under different starting conditions. Its developers’ Rule Inference for Transfer (RIFT) method improved prediction accuracy and physical consistency, including when rules were rare or data distributions changed. Separately, Google says its Willow quantum processor won the four-qubit magic square game 98.3% of the time, above the 88.89% classical limit. Google also reported results from tests involving up to 71 and 105 qubits, and said structured games can expose hardware errors that randomized tests may miss.

Other developments span AI risk, education, maritime research and regulation. Anthropic chief Dario Amodei has warned that malicious actors could use AI to help make bioweapons and that super-intelligent AI could pose an existential threat; the article also raises the possibility of investor losses if an AI bubble bursts. In South Korea, 111 students were caught with prohibited devices during the Suneung college entrance exam, following concerns about AI-assisted cheating; a separate report says exam candidate Lee Jin-woo, whose perfect score had drawn scrutiny, died of cardiac arrest on exam day. Researchers have also published an open-source workflow for maritime AI using public ship-tracking AIS data and Transformer-based trajectory forecasting, while Singapore’s Monetary Authority is strengthening AI oversight expectations ahead of a major governance failure, according to a DAC Beachcroft lawyer.

Key Takeaways

  • Meta is reviewing its 2030 net-zero timeline and targets after 2025 emissions rose 45% to 11.8 million metric tonnes and electricity use increased 35%.
  • Meta has contracted for more than 30 gigawatts of clean power, but about 12 gigawatts is producing electricity; it says natural gas may be needed when utility supply is delayed.
  • Microsoft CEO Satya Nadella warned that AI-driven vulnerability discovery can leave companies too little time to test fixes safely.
  • CIS reported vulnerability disclosures rising from about 4,274 in October 2025 to 17,060 in October 2026, with roughly 90% not practically exploitable; governments face increased patching demands.
  • Saturna analyst Levi Stewart Zurbrugg identified AI supply-chain opportunities in Taiwan, the Netherlands, South Korea and Japan, while warning depreciation could curb S&P 500 earnings growth in 2027–28.
  • Zurbrugg said investors should monitor corporate debt and credit conditions; Saturna’s Islamic screen limits exposure to companies with debt above 30% of market capitalisation.
  • RIFT improved models’ physical-rule prediction on the TailWorld benchmark, including tests involving rare rules and changed data distributions.
  • Google says Willow achieved 98.3% success in a four-qubit magic square game, exceeding the 88.89% classical limit; tests with up to 71 and 105 qubits also examined performance beyond classical bounds.
  • Anthropic chief Dario Amodei has warned of AI-enabled bioweapon risks and potential existential danger from super-intelligent AI.
  • South Korea caught 111 students with banned devices during the Suneung; other reports describe an open-source AIS maritime forecasting workflow and Singapore’s strengthening AI oversight.

Meta reviews 2030 climate goal as AI infrastructure raises emissions

Meta is reconsidering the timeline and interim steps for its goal of reaching net zero emissions by 2030. The company reported that its net carbon equivalent emissions rose 45 percent in 2025 to 11.8 million tonnes, as it builds infrastructure for artificial intelligence. Meta set its goal in 2020 to cover emissions across its value chain, including purchased goods and equipment. Data centres are using more power, and the article notes that other major companies have also stepped back from climate pledges.

Meta reconsiders 2030 climate promise as AI emissions rise

Meta is reviewing its 2030 net zero goal, interim milestones and near term targets as it expands its artificial intelligence infrastructure. Its greenhouse gas emissions rose 45 percent in 2025 to 11.8 million metric tons, while electricity use increased 35 percent. Meta says most of the emissions increase came from its supply chain, including equipment and data centre construction. The company has contracted for more than 30 gigawatts of clean and renewable power, but about 12 gigawatts is producing electricity so far. Meta says it may use natural gas plants when utilities cannot provide power quickly enough.

Meta rethinks 2030 net zero goal as AI data centres add emissions

Meta executives are reconsidering the company’s goal of reaching net zero emissions by 2030, according to its 2026 sustainability report. The report says the company’s operating environment has changed as it builds AI data centres. It describes renewable energy use and energy efficient data centres as steps Meta is taking to reduce its carbon footprint, while noting that AI data centres remain a source of emissions. The report says Meta remains committed to net zero by 2030 but must take further action.

Meta reviews climate goals after AI emissions rise 45 percent

Meta is reevaluating its climate goal after a report found that AI’s carbon footprint rose 45 percent over the past year. The report says the growth of AI is outpacing efficiency gains, as the technology becomes more powerful and more widely used. It points to increased use in fields such as healthcare and finance as a driver of growth. The report warns that rising AI energy use could make it harder for technology companies to meet their climate goals. Meta did not respond to a request for comment.

Saturna analyst weighs AI investment risks and global opportunities

Saturna Capital senior investment analyst and portfolio manager Levi Stewart Zurbrugg discussed global stocks, AI supply chains, interest rates and the risks of rising spending on AI infrastructure with Proactive’s Stephen Gunnion. He highlighted opportunities in Taiwan’s semiconductor industry, Dutch lithography technology, Korean memory manufacturing and Japanese optical fibre suppliers. Zurbrugg warned that depreciation costs from hyperscaler spending could significantly offset S&P 500 earnings growth in 2027 and 2028. He also said investors should watch corporate debt and credit conditions as companies rely more on outside financing. Saturna’s Islamic investment screening limits exposure to companies whose debt exceeds 30 percent of market capitalisation, and Zurbrugg discussed international diversification through the Saturna Global Equity ETF.

Saturna analyst sees AI opportunities abroad and warns of spending risks

Saturna Capital senior investment analyst and portfolio manager Levi Stewart Zurbrugg spoke with Proactive’s Stephen Gunnion about global investment opportunities and risks tied to AI spending. He pointed to Taiwan’s semiconductor industry, Dutch lithography technology, Korean memory makers and Japanese optical fibre suppliers. Zurbrugg warned that depreciation costs from hyperscalers’ AI investments could reduce S&P 500 earnings growth in 2027 and 2028, while greater reliance on outside financing makes corporate debt and credit conditions important to watch. He also described Saturna’s Islamic investment screen, which limits investment in companies with debt above 30% of market capitalisation, and discussed international diversification through the Saturna Global Equity ETF.

AI finds security flaws faster than companies can safely fix them

Microsoft CEO Satya Nadella warned that AI tools are creating a prolonged period of instability for companies. These tools can help find security flaws, but companies may struggle to put fixes in place safely. As a result, they may have less time to test updates before deploying them across critical systems.

RIFT helps AI models predict rare physical rules from video evidence

Researchers introduced TailWorld, a benchmark that tests whether models can learn a physical rule from object motion and apply it from a different starting point. The benchmark controls both how often each rule appears in training and how informative the video evidence is. The researchers also developed Rule Inference for Transfer, or RIFT, which learns how rules affect motion and uses observed interactions to identify the rule in a new case. Tests on TailWorld found that RIFT improved prediction accuracy and physical consistency, including when rare rules or changes in data distribution challenged the models.

Anthropic chief has warned of severe risks from artificial intelligence

The article points to Anthropic chief Dario Amodei’s warnings about the dangers of artificial intelligence. It says he has raised concerns about malicious actors using AI to make bioweapons and about super-intelligent AI potentially wiping out humanity. The piece argues that Anthropic’s new investors have been warned about these risks and could face losses if the AI bubble bursts.

Google’s Willow processor beats classical limits in quantum game tests

Google says its Willow processor won the four-qubit magic square game 98.3% of the time, above the classical limit of 88.89%. The team says hardware noise caused the remaining losses. Separate experiments included a multiplayer parity game using up to 71 qubits and a hidden linear function problem using 105 qubits, where results exceeded a conjectured bound for restricted classical circuits. Google also says these structured tasks revealed subtle hardware errors that randomized tests may hide, and researchers propose using them to assess quantum processors.

South Korea catches 111 students with banned devices during college exam

South Korea used emergency disaster alerts during its College Scholastic Ability Test, or Suneung, to catch 54 students with unauthorized devices. The first alert sounded ten minutes before the exam began and caught 31 students, whose scores were canceled; two later alerts caught 23 more. Officials reported that 111 students were caught overall, including 52 found through methods such as metal detectors, WiFi checks and physical inspections, and five caught using prohibited devices. The crackdown followed a tax accountant exam candidate’s confession to using AI and a micro-camera to cheat, prompting Prime Minister Lee Jae Myung to call for stronger action against AI-assisted cheating. The article also reports that LEET candidate Lee Jin-woo, whose perfect score had drawn scrutiny, was found dead of cardiac arrest on exam day, with a note saying he was mentally and physically exhausted.

AI driven vulnerability searches could mean more outages for governments

Cybersecurity experts at the Center for Internet Security warned that state and local governments should prepare for more frequent outages and maintenance as AI speeds up the discovery of software vulnerabilities. Randy Rose said disclosures rose from about 4,274 in October 2025 to 17,060 in October 2026, though about 90% are not practically exploitable. Many flaws affect commercial and open-source products that governments rely on, creating more patches for small security teams to track and install. Recent Microsoft outages cut some organizations off from Outlook, SharePoint and Teams, highlighting the need for backup communications plans. Rose said governments can reduce risk by following trusted security practices, while CIS also observed social engineering and typo-squatted domains during the recent disruptions.

Open source workflow aims to make maritime AI more reliable

A study presents a reproducible geospatial AI workflow for maritime analysis using public Automatic Identification System, or AIS, observations from multiple data providers. Built around damast, an open-source processing and machine-learning library from Simula Research Laboratory, the workflow includes metadata-aware data checks, reusable processing pipelines and maritime trajectory construction. Researchers demonstrate machine-learning experiments using a Transformer-based model to forecast vessel trajectories. They also describe a planned extension that would track data sources and data quality, and examine how those properties relate to model performance. The paper separates the capabilities shown in the current workflow from trustworthiness features proposed for future development.

Singapore strengthens oversight of artificial intelligence

Singapore has been developing its approach to responsible AI use for several years. New guidelines from the regulator reflect the growing scale and sophistication of AI adoption. Andrew Robinson of DAC Beachcroft said the Monetary Authority of Singapore appears to be setting expectations before a major incident reveals weaknesses in governance. The article is dated 20 November 2026.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI infrastructure data centers greenhouse gas emissions net zero renewable energy clean energy AI cybersecurity software vulnerabilities cybersecurity patching AI supply chain semiconductors AI investment corporate debt physical reasoning TailWorld benchmark quantum computing Google Willow AI safety AI-enabled bioweapons AI regulation

Comments

Loading...