Reshaping the Semiconductor Water Recovery Landscape: the Future of MLD and ZLD

In this article, UltraFacility caught up with Steven Lam from Gradiant to discuss all things water recovery in the semiconductor world. From membranes, to crystallizers, to smart monitoring tools, Steven shared Gradiant’s view on how chipmakers can reuse more, waste less, and balance cost with efficiency to get to the next frontier of wastewater recovery.

Why are minimum liquid discharge (MLD) and zero liquid discharge (ZLD) becoming increasingly popular options in the semiconductor industry?

MLD and ZLD are growing in popularity because there's growing attention from regional authorities and planners on how semiconductor fabs and data centers use water. For example, Amazon’s $3.6 billion data center project in Tucson was recently cancelled due to concerns over water and electricity usage. The city rejected it, citing sustainability concerns. That decision eliminated $250 million in tax revenue, 303,000 temporary construction jobs, and 180 permanent jobs. Even though data centers and semiconductor fabs use water differently, public scrutiny of one affects the other.

In Singapore, water is treated as a national security issue. The PUB agency monitors all water - from rainfall to industrial discharge. Semiconductors account for about 20% of non-domestic water demand in Singapore. While Singapore doesn’t mandate ZLD, it encourages recycling, and many companies are adopting MLD approaches.

As facilities aim for higher water recovery, total dissolved solids (TDS) in the reject stream may exceed discharge limits, and companies must switch to ZLD and incinerate the waste.

Brownfield sites tend to improve water recovery incrementally since they already have permits. Greenfield sites, however, face stricter scrutiny and are more likely to adopt ZLD from the outset. Geography plays a role too, as regions with scarce water resources push harder for these solutions before local authorities will issue permits.

Why would semiconductor manufacturers choose MLD over ZLD?

ZLD is significantly more expensive and offers only marginal gains in water recovery over MLD. Most recovery happens at low TDS levels, which are cheaper to treat. ZLD becomes viable when it offsets third-party disposal costs or meets regulatory requirements, like in India, where all new semiconductor facilities must implement ZLD.

Micron’s new fab in Boise is a good example of a strategic choice to implement a ZLD system. Micron reuses all industrial wastewater on-site, reducing reliance on city infrastructure and helping the City of Boise meet its phosphorus reduction goals. This makes it easier for the city to approve expansion without investing in additional infrastructure.

What are some emerging trends around the technologies that are being integrated into MLD and ZLD systems?

High-recovery membrane systems are becoming popular, and an obvious choice. They're smaller, lower in CAPEX and energy consumption than thermal technologies and they treat water efficiently. Traditional thermal evaporators are energy-intensive and expensive, as well as having a large footprint within the treatment facility. There is a general push to do as much as you can with membranes before switching to a downstream thermal evaporator and eventually a crystallizer.

Thermal evaporators haven’t disappeared from the flow scheme entirely. Companies are choosing to minimize but not eliminate their use. These boiler systems can still have some economic advantages when dealing with extremely high TDS levels. I think there’s still a level of hesitance to completely change the whole flow sheet before the industry has fully adopted these practices.

The industry is cautious, implementing new technologies in stages, but at the end of the day companies are willing to invest in innovation. They are constantly pushing to consume less water, use less energy, and they have the capital to invest to implement these changes.

Are there any technologies in development that could benefit MLD and ZLD systems in the future?

Yes. Currently, ZLD systems end with thermal crystallizers driven by either steam, gas or boiler systems which require a lot of energy. We are looking toward pressure-driven or electrochemical crystallizers, which could revolutionize flow schemes and reduce costs. This new crystallizer technology, crucially non thermally driven, could open new solutions. If we can switch to a pressure driven crystallization system, we will see a paradigm shift like what is happening in MLD, where folks are replacing thermal evaporators with membrane systems. It’s happening faster every year – in a few years I think you may no longer see thermal MLD systems at all.

Once you can make thermal crystallizers obsolete, the flow scheme possibilities are endless.

Do you have any products using pressure-driven crystallization?

Gradiant is looking into this, but it’s still an emerging technology so it’s still in early development and has only been used in specific applications. It’s technically possible—we’re working on making it practical for broader use.

Are smart monitoring tools being incorporated into MLD and ZLD systems?

Yes, especially for maintenance. Gradiant integrates advanced monitoring tools into its SmartOps AI solution for water and wastewater facilities, enabling real-time performance monitoring and optimization. Smart monitoring tools are increasingly used for predictive maintenance of pumps and equipment, helping maintain up time without adding capital costs. Semiconductor fabs can’t afford downtime, so it’s important to increase redundancy where you can without increasing CapEx. Software with sensors makes decision-making about predictive maintenance easier for operators. We are also using soft sensors to predict unmeasured parameters. When pursuing such high recovery, the risks of contaminants fouling the membranes or treatment systems increase. Using a physics-based approach in combination with an understanding of the water matrix, you can characterize water without implementing sensors for every parameter. For example, if TDS levels increase you can infer that other contaminants are also going up. It gives you a third parameter that you can use in your projection tool which helps you modulate how much recovery you’re getting based on water quality, or other OpEx factors.

How else can manufacturers balance cost and efficiency?

Cost is always critical – facilities are reluctant to implement tech that will be much more costly than their existing operations. We aim to improve efficiency without increasing CapEx or OpEx. Technologies must offer efficiency gains that are more than marginal in order to persuade the end-user. One area with enormous potential is recycling wastewater to the front end of UPW or cooling water treatment. Semiconductor wastewater is high quality due to strict discharge limits and could be reused - it’s low-cost to treat to product level and it helps to reduce intake and discharge. But it’s not yet widely adopted and embraced as best practice yet – even though the economics make sense, the practice needs momentum to overcome reluctance in adopting it.

So, it’s about building momentum and demonstrating proof of concept?

Exactly. Champions like TSMC and Micron lead the way by adopting best practices which then influence smaller fabs. In Taiwan, for example, many companies follow TSMC’s lead because they trust that TSMC has evaluated the efficiency and economics and proven that a practice is worthwhile. Once a major player proves a solution works, others will then make the jump to adopt it. The major companies take small incremental steps to improve efficiency, while smaller companies take big steps over time to catch up.

What choices must be considered when balancing sustainability parameters like water consumption, energy use, emissions, and waste disposal?

You need to improve one metric without negatively impacting others, and again incremental improvements are key to this. There are hard KPIs that facilities must meet, such as discharge standards, but everything else is about trying to make holistic improvements in a stepwise fashion. It means you must be willing to iterate more, to take on challenges repeatedly but it shouldn’t fall to the semiconductor company to make all the sustainability decisions. Semiconductor companies and other large corporations should focus on deliverables such as profitability and efficiency. It needs to be the municipalities that understand the local needs, like water scarcity or pollution, to guide sustainability.

Can materials from MLD and ZLD waste be recovered and reused?

Absolutely. Many of the raw materials that go into chipmaking can be recycled back into the process or repurposed downstream. For example, copper and silicon can be reused in other production facilities. Calcium Fluoride can be captured and directed into construction and heavy industry, while IPA and other solvents can also be reclaimed for reused in cleaning and process steps. By turning waste into a resource, MLD and ZLD systems not only advance circular economy goals but also help semiconductor fabs achieve zero liquid discharge while reducing operating costs.