Coinbase (NASDAQ: COIN) is leveraging artificial intelligence (AI) to tackle the recurring issue of its digital asset exchange going offline during volatile market periods. Historically, Coinbase has been notorious for extended outages whenever the crypto market experiences sudden surges or abrupt declines in the fiat value of major tokens. This has often led to significant financial losses for its users who find themselves unable to buy or sell their tokens or close out derivative positions.

On August 26, Coinbase announced a major breakthrough: the development and deployment of a machine learning model designed to predict spikes in user traffic. This model automatically scales databases to prevent downtime and enhance platform efficiency. By optimizing infrastructure costs and avoiding over-provisioning, Coinbase aims to maintain reliability even during unpredictable market movements.

Coinbase’s reputation has suffered due to frequent outages and questionable “planned maintenance” announcements that coincide with system investigations. A notable incident occurred on March 11, 2023, when Coinbase paused USDC stablecoin conversions to dollars during a weekend bank closure. In reality, Silicon Valley Bank, where USDC issuer Circle held $3.3 billion in fiat reserves, failed after a bank run, leaving Circle without dollars to convert until the Federal Reserve intervened.

Such outages have led to widespread criticism and suspicion. Some customers accuse Coinbase of deliberately going offline to act as a brake on downward trajectories or to cover internal issues before customers can react. Additionally, Coinbase’s decision to cut its ‘Website Hosting & Infrastructure’ spending from $440 million in 2022 to $192 million in 2023 while continuing to award nearly $1 billion annually in stock-based compensation to executives has raised eyebrows. This decision calls into question Coinbase’s commitment to maintaining robust infrastructure for customer access.

The new AI model represents a shift in strategy. Traditional methods of handling traffic surges were reactive and often too late. The AI model is designed to anticipate traffic spikes and scale resources proactively, aiming for at least an hour’s lead time. By incorporating external signals such as fluctuations in the fiat prices of BTC, ETH, and other prominent tokens, the model attempts to predict traffic thresholds more accurately.

However, the AI system is not foolproof. It may miss traffic spikes or scale up based on false predictions, which could result in unnecessary costs. To balance this, Coinbase designed the system to scale 2x its current level at the sign of a spike but “de-escalate” if the traffic level does not meet the predicted target within six hours.

A potential flaw in the AI model is its susceptibility to ‘hallucination,’ where it makes decisions based on erroneous information. Misinterpreting rising traffic as falling traffic or assuming all is well due to misleading public assertions could lead to further issues.

Coinbase’s focus on AI might have contributed to its delayed attention to other projects like Base, its Ethereum Layer 2 network. Launched just over a year ago, Base has quickly garnered significant user activity, with over one million daily active users and 4.6 million weekly active users as of August 24. Much of this activity revolves around the nearly 1.4 million different memecoins launched on Base since March 1. Despite their dubious value, these tokens have driven transaction volumes, making Base the leading Ethereum Layer 2 in terms of transactions since July.

Base users currently pay around $100,000 in daily transaction fees, of which only a small fraction goes to the Ethereum main layer. As the sole sequencer or transaction processor for Base, Coinbase benefits directly from these fees.

In February 2023, Coinbase promised gradual decentralization of Base. However, as of August, CEO Brian Armstrong’s statement that “others can run Base’s sequencers over time” remains unfulfilled, raising doubts about the timeline.

Coinbase has been promoting Base as a comprehensive solution for all crypto needs, including using USDC as the primary payment method and onboarding users via Coinbase Wallet. This approach aims to create a more proprietary version of BTC’s Lightning Network on Ethereum.

On August 13, Coinbase teased a new initiative called “cbBTC,” suggesting plans to introduce its own ‘wrapped’ BTC on Base. Wrapped BTC (WBTC) is an Ethereum token pegged 1:1 to BTC’s value, enabling users to leverage Ethereum’s smart contracts. WBTC, launched by BitGo in 2019, already has a market cap exceeding $9 billion and is available on Base. However, Coinbase’s cbBTC would give Armstrong more control over this wrapped asset.

Coinbase’s cbBTC plans coincided with BitGo’s strategic partnership with BiT Global, backed by Tron founder Justin Sun. Given Sun’s controversial history and questionable asset management practices, this partnership has raised concerns, leading MakerDAO to disable future WBTC borrows.

In summary, while Coinbase is making strides with AI and expanding its Base network, it must address infrastructure reliability and transparency concerns. For AI to work effectively and legally within an enterprise blockchain system, it needs robust data input quality and ownership measures to ensure data safety and immutability.

For more insights into emerging technologies like Enterprise blockchain and their impact on AI, explore CoinGeek’s extensive coverage.

Watch: Teranode & the Web3 world with edge-to-edge electronic value system